Compare commits

..

132 Commits

Author SHA1 Message Date
Lei, HUANG
a8630cdb38 fix: clippy errors 2022-12-15 18:12:05 +08:00
Ruihang Xia
0f3dcc1b38 fix: Fix All The Tests! (#752)
* fix: Fix several tests compile errors

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix: some compile errors in tests

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix: compile errors in frontend tests

* fix: compile errors in frontend tests

* test: Fix tests in api and common-query

* test: Fix test in sql crate

* fix: resolve substrait error

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* chore: add more test

* test: Fix tests in servers

* fix instance_test

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* test: Fix tests in tests-integration

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
Co-authored-by: Lei, HUANG <mrsatangel@gmail.com>
Co-authored-by: evenyag <realevenyag@gmail.com>
2022-12-15 17:47:14 +08:00
evenyag
7c696dae08 Merge branch 'develop' into replace-arrow2 2022-12-15 15:29:35 +08:00
LFC
61d8bc2ea1 refactor(frontend): minor changes around FrontendInstance constructor (#748)
* refactor: minor changes in some testing codes

Co-authored-by: luofucong <luofucong@greptime.com>
2022-12-15 14:34:40 +08:00
Yingwen
142dee41d6 fix: Fix compiler errors in script crate (#749)
* fix: Fix compiler errors in state.rs

* fix: fix compiler errors in state

* feat: upgrade sqlparser to 0.26

* fix: fix datafusion engine compiler errors

* fix: Fix some tests in query crate

* fix: Fix all warnings in tests

* feat: Remove `Type` from timestamp's type name

* fix: fix query tests

Now datafusion already supports median, so this commit also remove the
median function

* style: Fix clippy

* feat: Remove RecordBatch::pretty_print

* chore: Address CR comments

* feat: Add column_by_name to RecordBatch

* feat: modify select_from_rb

* feat: Fix some compiler errors in vector.rs

* feat: Fix more compiler errors in vector.rs

* fix: fix table.rs

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix: Fix compiler errors in coprocessor

* fix: Fix some compiler errors

* fix: Fix compiler errors in script

* chore: Remove unused imports and format code

* test: disable interval tests

* test: Fix test_compile_execute test

* style: Fix clippy

* feat: Support interval

* feat: Add RecordBatch::columns and fix clippy

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
Co-authored-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-15 14:20:35 +08:00
Ruihang Xia
e3785fca70 docs: change logo in readme automatically based on github theme (#743)
* docs: adaptive logo on theme

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* switch logos

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* aligh center

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* adjust stylet

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* use new logo image

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-14 19:32:51 +08:00
Lei, HUANG
ce6d1cb7d1 fix: frontend compile errors (#747)
fix: fix compile errors in frontend
2022-12-14 18:30:16 +08:00
Yingwen
dbb3034ecb fix: Fix compiler errors in query crate (#746)
* fix: Fix compiler errors in state.rs

* fix: fix compiler errors in state

* feat: upgrade sqlparser to 0.26

* fix: fix datafusion engine compiler errors

* fix: Fix some tests in query crate

* fix: Fix all warnings in tests

* feat: Remove `Type` from timestamp's type name

* fix: fix query tests

Now datafusion already supports median, so this commit also remove the
median function

* style: Fix clippy

* feat: Remove RecordBatch::pretty_print

* chore: Address CR comments

* Update src/query/src/query_engine/state.rs

Co-authored-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-14 17:42:07 +08:00
shuiyisong
fda9e80cbf feat: impl static_user_provider (#739)
* feat: add MemUserProvider and impl auth

* feat: impl user_provider option in fe and standalone mode

* chore: add file impl for mem provider

* chore: remove mem opts

* chore: minor change

* chore: refac pg server to use user_provider as indicator for using pwd auth

* chore: fix test

* chore: extract common code

* chore: add unit test

* chore: rebase develop

* chore: add user provider to http server

* chore: minor rename

* chore: change to ref when convert to anymap

* chore: fix according to clippy

* chore: remove clone on startcommand

* chore: fix cr issue

* chore: update tempdir use

* chore: change TryFrom to normal func while parsing anymap

* chore: minor change

* chore: remove to_lowercase
2022-12-14 16:38:29 +08:00
Lei, HUANG
756c068166 feat: logstore compaction (#740)
* feat: add benchmark for wal

* add bin

* feat: impl wal compaction

* chore: This reverts commit ef9f2326

* chore: This reverts commit 9142ec0e

* fix: remove empty files

* fix: failing tests

* fix: CR comments

* fix: Mark log as stable after writer applies manifest

* fix: some cr comments and namings

* chore: rename all stable_xxx to obsolete_xxx

* chore: error message
2022-12-14 16:15:29 +08:00
Lei, HUANG
652d59a643 fix: remove unwrap 2022-12-13 17:51:14 +08:00
Lei, HUANG
fa971c6513 fix: errors in optimzer 2022-12-13 17:44:37 +08:00
evenyag
36c929e1a7 fix: Fix imports in optimizer.rs 2022-12-13 17:27:44 +08:00
dennis zhuang
6a4e2e5975 feat: promql create and skeleton (#720)
* feat: adds promql crate

* feat: adds promql-parser dependency and rfc doc

* fix: dependencies order in servers crate

* fix: forgot error.rs

* fix: comment

* fix: license header

* fix: remove docs/rfc/20221207_promql.md
2022-12-13 17:08:22 +08:00
Ruihang Xia
a712382fba Merge pull request #745
* fix nyc-taxi and util

* Merge branch 'replace-arrow2' into fix-others

* fix substrait

* fix warnings and error in test
2022-12-13 16:59:28 +08:00
Yingwen
4b644aa482 fix: Fix compiler errors in catalog and mito crates (#742)
* fix: Fix compiler errors in mito

* fix: Fix compiler errors in catalog crate

* style: Fix clippy

* chore: Fix use
2022-12-13 15:53:55 +08:00
Lei, HUANG
9ad6ddb26e fix: remove useless metaclient field from datanode Instance (#744) 2022-12-13 14:26:26 +08:00
Lei, HUANG
4defde055c feat: upgrade storage crate to arrow and parquet offcial impl (#738)
* fix: compile erros

* fix: parquet reader and writer

* fix: parquet reader and writer

* fix: WriteBatch IPC encode/decode

* fix: clippy errors in storage subcrate

* chore: remove suspicious unwrap

* fix: some cr comments

* fix: CR comments

* fix: CR comments
2022-12-13 11:58:50 +08:00
fys
c5661ee362 feat: support http basic authentication (#733)
* feat: support http auth

* add some unit test and log

* fix

* cr

* remove unused #[derive(Clone)]
2022-12-13 10:44:33 +08:00
zyy17
9b093463cc feat: add Makefile to aggregate the commands that developers always use (#736)
* feat: add Makefile to aggregate the commands that developers always use

* refactor: add 'clean' and 'unit-test' target

* refactor: add sqlness-test target and modify some decriptions format

Signed-off-by: zyy17 <zyylsxm@gmail.com>
2022-12-12 13:03:49 +08:00
zyy17
61e0f1a11c refactor: add tls option in frontend cli options (#735)
* refactor: add tls option in frontend cli options

* fix: add 'Eq' trait for fixing clippy error

* fix: remove redundant clone

Signed-off-by: zyy17 <zyylsxm@gmail.com>
2022-12-12 10:02:17 +08:00
Ruihang Xia
95b2d8654f fix: pre-cast to avoid tremendous match arms (#734)
Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-09 17:20:03 +08:00
Ning Sun
249ebc6937 feat: update pgwire and refactor pg auth handler (#732) 2022-12-09 17:01:55 +08:00
Ruihang Xia
42fdc7251a fix: Fix common grpc expr (#730)
* fix compile errors

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* rename fn names

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix styles

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix wranings in common-time

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-09 14:24:04 +08:00
elijah
c1b8981f61 refactor(mito): change the table path to schema/table_id (#728)
refactor: change the table path to `schema/table_id`
2022-12-09 12:59:16 +08:00
Jiachun Feng
949cd3e3af feat: move_value & delete_route (#707)
* feat: move_value & delete_route

* chore: minor refactor

* chore: refactor unit test of metaclient

* chore: map to kv

* Update src/meta-srv/src/service/router.rs

Co-authored-by: Yingwen <realevenyag@gmail.com>

* Update src/meta-srv/src/service/router.rs

Co-authored-by: Yingwen <realevenyag@gmail.com>

* chore: by code review

Co-authored-by: Yingwen <realevenyag@gmail.com>
2022-12-09 11:07:48 +08:00
SSebo
b26982c5d7 feat: support timestamp new syntax (#697)
* feat: support timestamp new syntax

* fix: not null at end of new time stamp index syntax

* chore: simplify code
2022-12-09 10:52:14 +08:00
Ruihang Xia
d0892bf0b7 fix: Fix compile error in server subcrate (#727)
* fix: Fix compile error in server subcrate

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* remove unused type alias

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* explicitly panic

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* Update src/storage/src/sst/parquet.rs

Co-authored-by: Yingwen <realevenyag@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
Co-authored-by: Yingwen <realevenyag@gmail.com>
2022-12-08 20:27:53 +08:00
Ruihang Xia
fff530cb50 fix common record batch
Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-08 17:58:53 +08:00
Yingwen
b936d8b18a fix: Fix common::grpc compiler errors (#722)
* fix: Fix common::grpc compiler errors

This commit refactors RecordBatch and holds vectors in the RecordBatch
struct, so we don't need to cast the array to vector when doing
serialization or iterating the batch.

Now we use the vector API instead of the arrow API in grpc crate.

* chore: Address CR comments
2022-12-08 17:51:20 +08:00
Lei, HUANG
1bde1ba399 fix: row group pruning (#725)
* fix: row group pruning

* chore: use macro to simplify stats implemetation

* fxi: CR comments

* fix: row group metadata length mismatch

* fix: simplify code
2022-12-08 17:44:04 +08:00
Ruihang Xia
3687bc7346 fix: Fix tests and clippy for common-function subcrate (#726)
* further fixing

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix all compile errors in common function

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix tests

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix clippy

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* revert test changes

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-08 17:01:54 +08:00
fys
4fdf26810c feat: support auth in frontend (#688)
* feat: add UserProvider trait

* chore: minor fix

* support pg mysql

* refactor and add some logs

* chore: add license

Co-authored-by: shuiyisong <xixing.sys@gmail.com>
2022-12-08 11:51:52 +08:00
Ruihang Xia
587bdc9800 fix: fix other compile error in common-function (#719)
* further fixing

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix all compile errors in common function

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-08 11:38:07 +08:00
dennis zhuang
7f59758e69 feat: bump opendal version to 0.22 (#721)
* feat: bump opendal version to 0.22

* fix: LoggingLayer
2022-12-08 11:19:21 +08:00
Yingwen
58c26def6b fix: fix argmin/percentile/clip/interp/scipy_stats_norm_pdf errors (#718)
fix: fix argmin/percentile/clip/interp/scipy_stats_norm_pdf compiler errors
2022-12-07 19:55:07 +08:00
Ruihang Xia
6f3baf96b0 fix: fix compile error for mean/polyval/pow/interp ops (#717)
* fix: fix compile error for mean/polyval/pow/interp ops

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* simplify type bounds

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-07 16:38:43 +08:00
Yingwen
a898f846d1 fix: Fix compiler errors in argmax/rate/median/norm_cdf (#716)
* fix: Fix compiler errors in argmax/rate/median/norm_cdf

* chore: Address CR comments
2022-12-07 15:28:27 +08:00
Ruihang Xia
a562199455 Revert "fix: fix compile error for mean/polyval/pow/interp ops"
This reverts commit fb0b4eb826.
2022-12-07 15:13:58 +08:00
Ruihang Xia
fb0b4eb826 fix: fix compile error for mean/polyval/pow/interp ops
Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-07 15:12:28 +08:00
Zheming Li
a521ab5041 fix: set default value when fail to get git info instead of panic (#696)
fix: set default value when fail to git info instead of panic
2022-12-07 13:16:27 +08:00
LFC
833216d317 refactor: directly invoke Datanode methods in standalone mode (part 1) (#694)
* refactor: directly invoke Datanode methods in standalone mode

* test: add more unit tests

* fix: get rid of `println` in testing codes

* fix: resolve PR comments

* fix: resolve PR comments

Co-authored-by: luofucong <luofucong@greptime.com>
2022-12-07 11:37:59 +08:00
Yingwen
2ba99259e1 feat: Implements diff accumulator using WrapperType (#715)
* feat: Remove usage of opaque error from common::recordbatch

* feat: Remove opaque error from common::query

* feat: Fix diff compiler errors

Now common_function just use common_query's Error and Result. Adds
a LargestType associated type to LogicalPrimitiveType to get the largest
type a logical primitive type can cast to.

* feat: Remove LargestType from NativeType trait

* chore: Update comments

* feat: Restrict Scalar::RefType of WrapperType to itself

Add trait bound `for<'a> Scalar<RefType<'a> = Self>` to WrapperType

* chore: Address CR comments

* chore: Format codes
2022-12-07 11:13:24 +08:00
Ruihang Xia
551cde23b1 Merge branch 'dev' into replace-arrow2 2022-12-07 10:50:27 +08:00
Ruihang Xia
90c832b33d refactor: drop support of physical plan query interface (#714)
* refactor: drop support of physical plan query interface

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* refactor: collapse server/grpc sub-module

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* refactor: remove unused errors

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-06 19:23:32 +08:00
LFC
8959dbcef8 feat: Substrait logical plan (#704)
* feat: use Substrait logical plan to query data from Datanode in Frontend in distributed mode

* fix: resolve PR comments

* fix: resolve PR comments

* fix: resolve PR comments

Co-authored-by: luofucong <luofucong@greptime.com>
2022-12-06 19:21:57 +08:00
Yingwen
653906d4fa fix: Fix common::query compiler errors (#713)
* feat: Move conversion to ScalarValue to value.rs

* fix: Fix common::query compiler errors

This commit also make InnerError pub(crate)
2022-12-06 16:45:54 +08:00
Ruihang Xia
829ff491c4 fix: common-query subcrate (#712)
* fix: record batch adapter

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix error enum

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-06 16:32:52 +08:00
Yingwen
b32438e78c feat: Fix some compiler errors in common::query (#710)
* feat: Fix some compiler errors in common::query

* feat: test_collect use vectors api
2022-12-06 15:32:12 +08:00
Lei, HUANG
0ccb8b4302 chore: delete datatypes based on arrow2 2022-12-06 15:01:57 +08:00
Lei, HUANG
b48ae21b71 fix: api crate (#708)
* fix: rename ConcreteDataType::timestamp_millis_type to ConcreteDataType::timestamp_millisecond_type. fix other warnings regarding timestamp

* fix: revert changes in datatypes2

* fix: helper
2022-12-06 14:56:59 +08:00
discord9
2034b40f33 chore: update RustPython dependence(With a tweaked fork) (#655)
* refactor: update RsPy

* depend: add `rustpython-pylib`

* feat: add_frozen stdlib for every vm init

* feat: limit stdlib to a selected few

* chore: use `rev` instead of branch` im depend

* refactor: rename to allow_list

* feat: use opt level one

* doc: add username for TODO&change optimize to 0

* style: fmt .toml
2022-12-06 14:15:00 +08:00
evenyag
3c0adb00f3 feat: Fix recordbatch test compiling issue 2022-12-06 12:03:06 +08:00
evenyag
8c66b7d000 feat: Fix common::recordbatch compiler errors 2022-12-06 11:55:19 +08:00
evenyag
99371fd31b chore: sort Cargo.toml 2022-12-06 11:39:15 +08:00
evenyag
fe505fecfd feat: Make recordbatch compile 2022-12-06 11:38:59 +08:00
SSebo
55e6be7af1 fix: test_server_require_secure_client_secure (#701) 2022-12-06 10:38:54 +08:00
discord9
f9bfb121db feat: add rate() udf (#508)
* feat: rewrite `rate` UDF

* feat: rename to `prom_rate`

* refactor: solve conflict&add license

* refactor: import arrow
2022-12-06 10:30:13 +08:00
evenyag
cc1ec26416 feat: Switch to datatypes2 2022-12-05 20:30:47 +08:00
Ruihang Xia
504059a699 chore: fix wrong merge commit
Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-05 20:11:22 +08:00
Ruihang Xia
7151deb4ed Merge branch 'dev' into replace-arrow2 2022-12-05 20:10:37 +08:00
Ruihang Xia
6fb413ae50 ci: add toml format linter (#706)
* chore: run taplo format

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* ci: add workflow to check toml

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* rerun formatter with ident to 4 spaces

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* update check command

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-12-05 20:03:10 +08:00
Ruihang Xia
beb07fc895 feat: new datatypes subcrate based on the official arrow (#705)
* feat: Init datatypes2 crate

* chore: Remove some unimplemented types

* feat: Implements PrimitiveType and PrimitiveVector for datatypes2 (#633)

* feat: Implement primitive types and vectors

* feat: Implement a wrapper type

* feat: Remove VectorType from ScalarRef

* feat: Move some trait bound from NativeType to WrapperType

* feat: pub use  primitive vectors and builders

* feat: Returns error in try_from when type mismatch

* feat: Impl PartialEq for some vectors

* test: Pass vector tests

* chore: Add license header

* test: Pass more vector tests

* feat: Implement some methods of vector Helper

* test: Pass more tests

* style: Fix clippy

* chore: Add license header

* feat: Remove IntoValueRef trait

* feat: Add NativeType trait bound to WrapperType::Native

* docs: Explain what is wrapper type

* chore: Fix typos

* refactor: LogicalPrimitiveType::type_name returns str

* feat: Implements DateType and DateVector (#651)

* feat: Implement DateType and DateVector

* test: Pass more value and data type tests

* chore: Address CR comments

* test: Skip list value test

* feat: datatypes2 datetime (#661)

* feat: impl DateTime type and vector

* fix: add license header

* fix: CR comments and add more tests

* fix: customized serialization for wrapper type

* feat: Implements NullType and NullVector (#658)

* feat: Implements NullType and NullVector

* chore: Address CR comment

Co-authored-by: Ruihang Xia <waynestxia@gmail.com>

* chore: Address CR comment

Co-authored-by: Ruihang Xia <waynestxia@gmail.com>

* feat: Implements StringType and StringVector (#659)

* feat: implement string vector

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* add more test and from

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix clippy

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* cover NUL

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* feat: impl datatypes2/timestamp (#686)

* feat: add timestamp datatype and vectors

* fix: cr comments and reformat code

* chore: add some tests

* feat: Implements ListType and ListVector (#681)

* feat: Implement ListType and ListVector

* test: Pass more tests

* style: Fix clippy

* chore: Fix comment

* chore: Address CR comments

* feat: impl constant vector (#680)

* feat: impl constant vector

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix tests

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* Apply suggestions from code review

Co-authored-by: Yingwen <realevenyag@gmail.com>

* rename fn names

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* remove println

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
Co-authored-by: Yingwen <realevenyag@gmail.com>

* feat: Implements Validity (#684)

* feat: Implements Validity

* chore: remove pub from sub mod in vectors

* feat: Implements schema for datatypes2 (#695)

* feat: Add is_timestamp_compatible to DataType

* feat: Implement ColumnSchema and Schema

* feat: Impl RawSchema

* chore: Remove useless codes and run more tests

* chore: Fix clippy

* feat: Impl from_arrow_time_unit and pass schema tests

* chore: add more tests for timestamp (#702)

* chore: add more tests for timestamp

* chore: add replicate test for timestamps

* feat: Implements helper methods for vectors/values (#703)

* feat: Implement helper methods for vectors/values

* chore: Address CR comments

* chore: add more test for timestamp

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
Co-authored-by: evenyag <realevenyag@gmail.com>
Co-authored-by: Lei, HUANG <6406592+v0y4g3r@users.noreply.github.com>
Co-authored-by: Lei, HUANG <mrsatangel@gmail.com>
2022-12-05 19:59:23 +08:00
Ning Sun
4275e47bdb refactor: use updated mysql_async client (#698) 2022-12-05 11:18:32 +08:00
dennis zhuang
6720bc5f7c fix: validate create table request in mito engine (#690)
* fix: validate create table request in mito engine

* fix: comment

* chore: remove TIMESTAMP_INDEX in system.rs
2022-12-05 11:01:43 +08:00
Lei, HUANG
4052563248 fix: pr template task default state (#687) 2022-12-02 20:39:53 +08:00
dennis zhuang
952e1bd626 test: update dummy result (#693) 2022-12-02 19:22:37 +08:00
shuiyisong
8232015998 fix: cargo sort in pre-commit (#689) 2022-12-02 16:19:31 +08:00
Ruihang Xia
d82a3a7d58 feat: implement most of scalar function and selection conversion in substrait (#678)
* impl to_df_scalar_function

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* part of scalar functions

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* conjunction over filters

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* change the ser/de target to substrait::Plan

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* basic test coverage

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix typos and license header

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix clippy

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix CR comments

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* logs unsupported extension

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* Update src/common/substrait/src/df_expr.rs

Co-authored-by: Yingwen <realevenyag@gmail.com>

* address review comments

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* change format

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* replace context with with_context

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
Co-authored-by: Yingwen <realevenyag@gmail.com>
2022-12-02 14:46:05 +08:00
Ning Sun
0599465685 feat: inject current database/schema into query context for postgres protocol (#685)
* feat: inject current database/schema into query context

* test: avoid duplicate server setup
2022-12-02 11:49:39 +08:00
Mofeng
13d51250ba feat: add http /health api (#676)
* feat: add http `/health` api

* feat: add `/health` api test suit in http intergration test
2022-12-01 19:11:58 +08:00
LFC
6127706b5b feat: support "use" stmt part 1 (#672)
* feat: a bare sketch of session; support "use" in MySQL server; modify insertion and selection related codes in Datanode
2022-12-01 17:05:32 +08:00
dennis zhuang
2e17e9c4b5 feat: supports s3 storage (#656)
* feat: adds s3 object storage configuration

* feat: adds s3 integration test

* chore: use map

* fix: forgot license header

* fix: checking if bucket is empty in test_on

* chore: address CR issues

* refactor: run s3 test with dotenv

* chore: randomize grpc port for test

* fix: README in tests-integration

* chore: remove redundant comments
2022-12-01 10:59:14 +08:00
xiaomin tang
b0cbfa7ffb docs: add a roadmap link in README (#673)
* docs: add roadmap to README

* docs: missing period
2022-11-30 21:25:27 +08:00
Ruihang Xia
20172338e8 ci: Revert "ci: change CI unit test trigger" (#674)
Revert "ci: change CI unit test trigger (#671)"

This reverts commit 9c53f9b24c.
2022-11-30 21:23:40 +08:00
Ruihang Xia
9c53f9b24c ci: change CI unit test trigger (#671)
Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-11-30 20:19:35 +08:00
Dongxu Wang
6d24f7ebb6 refactor: bump axum 0.6, use recommended way to nest routes (#668) 2022-11-30 20:04:33 +08:00
SSebo
68c2de8e45 feat: mysql and pg server support tls (#641)
* feat: mysql and pg server support tls

* chore: replace opensrv-mysql to original

* chore: TlsOption is required but supply default value

* feat: mysql server support force tls

* chore: move TlsOption to servers

* test: mysql server disable / prefer / required tls mode

* test: pg server disable / prefer / required tls mode

* chore: add doc and remove no used code

* chore: add TODO and restore cargo linker config
2022-11-30 12:46:15 +08:00
Yingwen
a17dcbc511 chore: fix SequenceNotMonotonic error message (#664)
* chore: fix SequenceNotMonotonic error message

previous sequence should greater than or equal to given sequence

* Apply suggestions from code review

Co-authored-by: Ruihang Xia <waynestxia@gmail.com>
2022-11-30 11:58:43 +08:00
Ning Sun
53ab19ea5a ci: remove assignees which is causing error (#663) 2022-11-30 11:36:35 +08:00
Ning Sun
84c44cf540 ci: fix doc label task on forked repo (#654) 2022-11-30 11:23:15 +08:00
LFC
020b9936cd fix: correctly detach spawned mysql listener task (#657)
Co-authored-by: luofucong <luofucong@greptime.com>
2022-11-29 18:39:18 +08:00
Ning Sun
75dcf2467b refactor: add tests-integration module (#590)
* refactor: add integration-tests module

* Apply suggestions from code review

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

* test: move grpc module to tests-integration

* test: adapt new standalone mode

* test: improve http assertion

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
2022-11-29 16:28:58 +08:00
Ruihang Xia
eea5393f96 feat: UI improvement for integration test runner (#645)
* improve dir resolving and start up ordering

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* fix orphan process

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* Update tests/runner/src/util.rs, fix typo

Co-authored-by: Dongxu Wang <dongxu@apache.org>

* simplify logic via tokio timeout

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
Co-authored-by: Dongxu Wang <dongxu@apache.org>
2022-11-29 15:32:39 +08:00
Ning Sun
3d312d389d ci: add doc label support for pr too (#650) 2022-11-29 15:21:12 +08:00
dennis zhuang
fdc73fb52f perf: cache python interpreter in TLS (#649)
* perf: cache python interpreter when executing coprocessors

* test: speedup test_execute_script by reusing interpreter

* fix: remove comment

* chore: use get_or_insert_with instead
2022-11-29 14:41:37 +08:00
Ning Sun
2a36e26d19 ci: add action to create doc issue when change labelled (#648)
ci: add action to create doc issue when change labeled
2022-11-29 14:25:57 +08:00
Zheming Li
baef640fe3 feat: add --version command line option (#632)
* add version command line option

* use concat!

Co-authored-by: Ruihang Xia <waynestxia@gmail.com>

Co-authored-by: Ruihang Xia <waynestxia@gmail.com>
2022-11-28 17:07:17 +08:00
dennis zhuang
5fddb799f7 feat: enable atomic write for file object storage (#643)
* fix: remove opendal from catalog dependencies

* feat: enable atomic writing for fs service
2022-11-28 16:01:32 +08:00
Dongxu Wang
f372229b18 fix: append table id to table data dir (#640) 2022-11-28 10:53:13 +08:00
Xuanwo
4085fc7899 chore: Bump OpenDAL to v0.21.1 (#639)
* deps: Bump OpenDAL to v0.21.1

Signed-off-by: Xuanwo <github@xuanwo.io>

* Avoid using raw types when not needed

Signed-off-by: Xuanwo <github@xuanwo.io>

Signed-off-by: Xuanwo <github@xuanwo.io>
2022-11-27 10:18:39 +08:00
Ruihang Xia
30940e692a feat: impl DROP TABLE on memory catalog based standalone mode (#630)
* feat: implement drop table for standalone mode

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* update integration test

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

* enhancement test

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>

Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-11-25 11:53:46 +08:00
Mike Yang
b371ce0f48 test: added tests for statements methods (#622)
* test: added tests for parse_column_default_constraint

* test: added test for sql_column_def_to_grpc_column_def

* refactor: remove hardcode in test
2022-11-25 11:35:06 +08:00
Lei, HUANG
ac7f52d303 fix: start datanode instance before frontend services (#634) 2022-11-25 11:25:57 +08:00
Dongxu Wang
051768b735 ci: add spell check with typos (#627) 2022-11-24 14:46:50 +08:00
fys
c5b0d2431f feat: remove InsertBatch in gRPC message (#570) 2022-11-24 14:04:48 +08:00
Lei, HUANG
4038dd4067 fix: add concurrency control for catalog manager (#619) 2022-11-24 11:10:33 +08:00
Dongxu Wang
8be0f05570 chore: able to config axum timeout in toml (#624) 2022-11-24 11:09:21 +08:00
zyy17
69f06eec8b ci: change scheduled release from nigthly to weekly (#623)
Signed-off-by: zyy17 <zyylsxm@gmail.com>
2022-11-24 11:05:35 +08:00
Ruihang Xia
7b37e99a45 feat: deregister table for MemoryCatalogManager (#620)
Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-11-24 09:36:27 +08:00
dennis zhuang
c09775d17f feat: adds metrics, tracing and retry layer to object-store (#621) 2022-11-23 11:40:03 +08:00
Francis Du
4a9cf49637 feat: support explain syntax (#546) 2022-11-22 21:22:32 +08:00
Ruihang Xia
9f865b50ab test: add dummy select case (#618)
Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-11-22 16:47:45 +08:00
Ruihang Xia
b407ebf6bb feat: integration test suite (#487)
Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-11-22 15:34:13 +08:00
Lei, HUANG
c144a1b20e feat: impl alter table in distributed mode (#572) 2022-11-22 15:17:25 +08:00
Ruihang Xia
d0686f9c19 Merge branch 'replace-arrow2' of github.com:GreptimeTeam/greptimedb into replace-arrow2 2022-11-21 17:43:40 +08:00
Ruihang Xia
221f3e9d2e Merge branch 'dev' into replace-arrow2 2022-11-21 17:42:15 +08:00
Yingwen
0791c65149 refactor: replace some usage of MutableBitmap by BitVec (#610) 2022-11-21 17:36:53 +08:00
evenyag
61c4a3691a chore: update dep of binary vector 2022-11-21 15:55:07 +08:00
evenyag
d7626fd6af feat: arrow_array switch to arrow 2022-11-21 15:39:41 +08:00
LFC
62fcb54258 fix: correctly open table when distributed datanode restart (#576)
Co-authored-by: luofucong <luofucong@greptime.com>
2022-11-21 15:15:14 +08:00
Ruihang Xia
e3201a4705 chore: replace one last datafusion dep
Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-11-21 14:29:59 +08:00
Ruihang Xia
571a84d91b chore: kick off. change datafusion/arrow/parquet to target version
Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-11-21 14:19:39 +08:00
Lei, HUANG
2b6b979d5a fix: remove datanode mysql options in standalone mode (#595) 2022-11-21 14:15:47 +08:00
Dongxu Wang
b6fa316c65 chore: correct typos (#589) (#592) 2022-11-21 14:07:45 +08:00
Lei, HUANG
ca5734edb3 feat: disable mysql server on datande when running standalone mode (#593) 2022-11-21 12:12:26 +08:00
Mike Yang
5428ad364e fix: make nullable as default when alter table (#591) 2022-11-21 12:11:19 +08:00
zyy17
663c725838 fix: fix nightly build error and fix typo (#588)
Signed-off-by: zyy17 <zyylsxm@gmail.com>
2022-11-21 11:49:36 +08:00
zyy17
c94b544e4a ci: modify image registry in release.yml (#582)
Signed-off-by: zyy17 <zyylsxm@gmail.com>
2022-11-19 09:19:54 +08:00
Ruihang Xia
f465040acc feat: lazy evaluated record batch stream (#573)
Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
2022-11-18 21:42:10 +08:00
Yingwen
22ae983280 refactor: Use re-exported arrow mod from datatypes crate (#571) 2022-11-18 18:38:07 +08:00
Igor Morozov
e1f326295f feat: implement DESCRIBE TABLE (#558)
Also need to support describe table in other catalog/schema
2022-11-18 16:34:00 +08:00
aievl
6d762aa9dc feat: update mysql default listen port to 4406 (#568)
Co-authored-by: zhaozhenhang <zhaozhenhang@kuaishou.com>
2022-11-18 14:55:11 +08:00
Ruihang Xia
d4b09f69ab docs: specify protoc version requirement (#564)
Signed-off-by: Ruihang Xia <waynestxia@gmail.com>
Co-authored-by: Yingwen <realevenyag@gmail.com>
2022-11-18 14:36:25 +08:00
Xuanwo
1f0b39cc8d chore: Bump OpenDAL to v0.20 (#569)
Signed-off-by: Xuanwo <github@xuanwo.io>
2022-11-18 14:17:38 +08:00
zyy17
dee5ccec9e ci: add nightly build job (#565) 2022-11-18 11:48:29 +08:00
dennis zhuang
f8788273d5 feat: drop column for alter table (#562)
* feat: drop column for alter table

* refactor: rename RemoveColumns to DropColumns

* test: alter table

* chore: error msg

Co-authored-by: Ruihang Xia <waynestxia@gmail.com>

* fix: test_parse_alter_drop_column

Co-authored-by: Ruihang Xia <waynestxia@gmail.com>
2022-11-17 23:00:16 +08:00
jay
df465308cc current blog url response as 404, should be https://greptime.com/blogs/index (#561) 2022-11-17 21:24:04 +08:00
LFC
e7b4d2b9cd feat: Implement table_info() for DistTable (#536) (#557)
* feat: Implement `table_info()`` for `DistTable` (#536)

* Update src/catalog/src/error.rs

Co-authored-by: Yingwen <1405012107@qq.com>

Co-authored-by: luofucong <luofucong@greptime.com>
Co-authored-by: Yingwen <1405012107@qq.com>
2022-11-17 18:40:58 +08:00
discord9
bf408e3b96 Update README.md (#552)
Add RustPython's Acknowledgement
2022-11-17 14:15:43 +08:00
dennis zhuang
73e6e2e01b fix: split code and output in README (#549) 2022-11-17 12:54:02 +08:00
Lei, Huang
8faa6b0f09 refactor: start options (#545)
* refactor: config options for frontend/datanode/standalone

* chore: rename MetaClientOpts::metasrv_addr to MetaClientOpts::metasrv_addrs

* fix: clippy

* fix: change default meta-srv addr to 127.0.0.1:3002
2022-11-17 11:47:39 +08:00
427 changed files with 20789 additions and 12810 deletions

4
.env.example Normal file
View File

@@ -0,0 +1,4 @@
# Settings for s3 test
GT_S3_BUCKET=S3 bucket
GT_S3_ACCESS_KEY_ID=S3 access key id
GT_S3_ACCESS_KEY=S3 secret access key

View File

@@ -13,7 +13,7 @@ Please explain IN DETAIL what the changes are in this PR and why they are needed
## Checklist
- [] I have written the necessary rustdoc comments.
- [] I have added the necessary unit tests and integration tests.
- [ ] I have written the necessary rustdoc comments.
- [ ] I have added the necessary unit tests and integration tests.
## Refer to a related PR or issue link (optional)

View File

@@ -26,6 +26,13 @@ env:
RUST_TOOLCHAIN: nightly-2022-07-14
jobs:
typos:
name: Spell Check with Typos
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v2
- uses: crate-ci/typos@v1.0.4
check:
name: Check
if: github.event.pull_request.draft == false
@@ -42,6 +49,23 @@ jobs:
- name: Run cargo check
run: cargo check --workspace --all-targets
toml:
name: Toml Check
if: github.event.pull_request.draft == false
runs-on: ubuntu-latest
timeout-minutes: 60
steps:
- uses: actions/checkout@v3
- uses: dtolnay/rust-toolchain@master
with:
toolchain: ${{ env.RUST_TOOLCHAIN }}
- name: Rust Cache
uses: Swatinem/rust-cache@v2
- name: Install taplo
run: cargo install taplo-cli --version ^0.8 --locked
- name: Run taplo
run: taplo format --check --option "indent_string= "
# Use coverage to run test.
# test:
# name: Test Suite

25
.github/workflows/doc-issue.yml vendored Normal file
View File

@@ -0,0 +1,25 @@
name: Create Issue in docs repo on doc related changes
on:
issues:
types:
- labeled
pull_request_target:
types:
- labeled
jobs:
doc_issue:
if: github.event.label.name == 'doc update required'
runs-on: ubuntu-latest
steps:
- name: create an issue in doc repo
uses: dacbd/create-issue-action@main
with:
owner: GreptimeTeam
repo: docs
token: ${{ secrets.DOCS_REPO_TOKEN }}
title: Update docs for ${{ github.event.issue.title || github.event.pull_request.title }}
body: |
A document change request is generated from
${{ github.event.issue.html_url || github.event.pull_request.html_url }}

View File

@@ -2,6 +2,9 @@ on:
push:
tags:
- "v*.*.*"
schedule:
# At 00:00 on Monday.
- cron: '0 0 * * 1'
workflow_dispatch:
name: Release
@@ -9,6 +12,12 @@ name: Release
env:
RUST_TOOLCHAIN: nightly-2022-07-14
# FIXME(zyy17): Would be better to use `gh release list -L 1 | cut -f 3` to get the latest release version tag, but for a long time, we will stay at 'v0.1.0-alpha-*'.
SCHEDULED_BUILD_VERSION_PREFIX: v0.1.0-alpha
# In the future, we can change SCHEDULED_PERIOD to nightly.
SCHEDULED_PERIOD: weekly
jobs:
build:
name: Build binary
@@ -106,8 +115,32 @@ jobs:
- name: Download artifacts
uses: actions/download-artifact@v3
- name: Configure scheduled build version # the version would be ${SCHEDULED_BUILD_VERSION_PREFIX}-YYYYMMDD-${SCHEDULED_PERIOD}, like v0.1.0-alpha-20221119-weekly.
shell: bash
if: github.event_name == 'schedule'
run: |
buildTime=`date "+%Y%m%d"`
SCHEDULED_BUILD_VERSION=${{ env.SCHEDULED_BUILD_VERSION_PREFIX }}-$buildTime-${{ env.SCHEDULED_PERIOD }}
echo "SCHEDULED_BUILD_VERSION=${SCHEDULED_BUILD_VERSION}" >> $GITHUB_ENV
- name: Create scheduled build git tag
if: github.event_name == 'schedule'
run: |
git tag ${{ env.SCHEDULED_BUILD_VERSION }}
- name: Publish scheduled release # configure the different release title and tags.
uses: softprops/action-gh-release@v1
if: github.event_name == 'schedule'
with:
name: "Release ${{ env.SCHEDULED_BUILD_VERSION }}"
tag_name: ${{ env.SCHEDULED_BUILD_VERSION }}
generate_release_notes: true
files: |
**/greptime-*
- name: Publish release
uses: softprops/action-gh-release@v1
if: github.event_name != 'schedule'
with:
name: "Release ${{ github.ref_name }}"
files: |
@@ -145,12 +178,12 @@ jobs:
tar xvf greptime-linux-arm64.tgz
rm greptime-linux-arm64.tgz
- name: Login to GitHub Container Registry
- name: Login to UCloud Container Registry
uses: docker/login-action@v2
with:
registry: ghcr.io
username: ${{ github.actor }}
password: ${{ secrets.GITHUB_TOKEN }}
registry: uhub.service.ucloud.cn
username: ${{ secrets.UCLOUD_USERNAME }}
password: ${{ secrets.UCLOUD_PASSWORD }}
- name: Login to Dockerhub
uses: docker/login-action@v2
@@ -158,11 +191,20 @@ jobs:
username: ${{ secrets.DOCKERHUB_USERNAME }}
password: ${{ secrets.DOCKERHUB_TOKEN }}
- name: Configure scheduled build image tag # the tag would be ${SCHEDULED_BUILD_VERSION_PREFIX}-YYYYMMDD-${SCHEDULED_PERIOD}
shell: bash
if: github.event_name == 'schedule'
run: |
buildTime=`date "+%Y%m%d"`
SCHEDULED_BUILD_VERSION=${{ env.SCHEDULED_BUILD_VERSION_PREFIX }}-$buildTime-${{ env.SCHEDULED_PERIOD }}
echo "IMAGE_TAG=${SCHEDULED_BUILD_VERSION:1}" >> $GITHUB_ENV
- name: Configure tag # If the release tag is v0.1.0, then the image version tag will be 0.1.0.
shell: bash
if: github.event_name != 'schedule'
run: |
VERSION=${{ github.ref_name }}
echo "VERSION=${VERSION:1}" >> $GITHUB_ENV
echo "IMAGE_TAG=${VERSION:1}" >> $GITHUB_ENV
- name: Set up QEMU
uses: docker/setup-qemu-action@v2
@@ -179,5 +221,6 @@ jobs:
platforms: linux/amd64,linux/arm64
tags: |
greptime/greptimedb:latest
greptime/greptimedb:${{ env.VERSION }}
ghcr.io/greptimeteam/greptimedb:${{ env.VERSION }}
greptime/greptimedb:${{ env.IMAGE_TAG }}
uhub.service.ucloud.cn/greptime/greptimedb:latest
uhub.service.ucloud.cn/greptime/greptimedb:${{ env.IMAGE_TAG }}

4
.gitignore vendored
View File

@@ -18,6 +18,7 @@ debug/
# JetBrains IDE config directory
.idea/
*.iml
# VSCode IDE config directory
.vscode/
@@ -31,3 +32,6 @@ logs/
# Benchmark dataset
benchmarks/data
# dotenv
.env

View File

@@ -9,7 +9,7 @@ repos:
rev: e6a795bc6b2c0958f9ef52af4863bbd7cc17238f
hooks:
- id: cargo-sort
args: ["--workspace", "--print"]
args: ["--workspace"]
- repo: https://github.com/doublify/pre-commit-rust
rev: v1.0

2963
Cargo.lock generated

File diff suppressed because it is too large Load Diff

View File

@@ -11,11 +11,11 @@ members = [
"src/common/function",
"src/common/function-macro",
"src/common/grpc",
"src/common/grpc-expr",
"src/common/query",
"src/common/recordbatch",
"src/common/runtime",
"src/common/substrait",
"src/common/insert",
"src/common/telemetry",
"src/common/time",
"src/datanode",
@@ -24,15 +24,19 @@ members = [
"src/log-store",
"src/meta-client",
"src/meta-srv",
"src/mito",
"src/object-store",
"src/promql",
"src/query",
"src/script",
"src/servers",
"src/session",
"src/sql",
"src/storage",
"src/store-api",
"src/table",
"src/mito",
"tests-integration",
"tests/runner",
]
[profile.release]

67
Makefile Normal file
View File

@@ -0,0 +1,67 @@
IMAGE_REGISTRY ?= greptimedb
IMAGE_TAG ?= latest
##@ Build
.PHONY: build
build: ## Build debug version greptime.
cargo build
.PHONY: release
release: ## Build release version greptime.
cargo build --release
.PHONY: clean
clean: ## Clean the project.
cargo clean
.PHONY: fmt
fmt: ## Format all the Rust code.
cargo fmt --all
.PHONY: docker-image
docker-image: ## Build docker image.
docker build --network host -f docker/Dockerfile -t ${IMAGE_REGISTRY}:${IMAGE_TAG} .
##@ Test
.PHONY: unit-test
unit-test: ## Run unit test.
cargo test --workspace
.PHONY: integration-test
integration-test: ## Run integation test.
cargo test integration
.PHONY: sqlness-test
sqlness-test: ## Run sqlness test.
cargo run --bin sqlness-runner
.PHONY: check
check: ## Cargo check all the targets.
cargo check --workspace --all-targets
.PHONY: clippy
clippy: ## Check clippy rules.
cargo clippy --workspace --all-targets -- -D warnings -D clippy::print_stdout -D clippy::print_stderr
.PHONY: fmt-check
fmt-check: ## Check code format.
cargo fmt --all -- --check
##@ General
# The help target prints out all targets with their descriptions organized
# beneath their categories. The categories are represented by '##@' and the
# target descriptions by '##'. The awk commands is responsible for reading the
# entire set of makefiles included in this invocation, looking for lines of the
# file as xyz: ## something, and then pretty-format the target and help. Then,
# if there's a line with ##@ something, that gets pretty-printed as a category.
# More info on the usage of ANSI control characters for terminal formatting:
# https://en.wikipedia.org/wiki/ANSI_escape_code#SGR_parameters
# More info on the awk command:
# https://linuxcommand.org/lc3_adv_awk.php
.PHONY: help
help: ## Display help messages.
@awk 'BEGIN {FS = ":.*##"; printf "\nUsage:\n make \033[36m<target>\033[0m\n"} /^[a-zA-Z_0-9-]+:.*?##/ { printf " \033[36m%-20s\033[0m %s\n", $$1, $$2 } /^##@/ { printf "\n\033[1m%s\033[0m\n", substr($$0, 5) } ' $(MAKEFILE_LIST)

View File

@@ -1,7 +1,12 @@
<p align="center">
<img src="/docs/logo-text-padding.png" alt="GreptimeDB Logo" width="400px"></img>
<picture>
<source media="(prefers-color-scheme: light)" srcset="/docs/logo-text-padding.png">
<source media="(prefers-color-scheme: dark)" srcset="/docs/logo-text-padding-dark.png">
<img alt="GreptimeDB Logo" src="/docs/logo-text-padding.png" width="400px">
</picture>
</p>
<h3 align="center">
The next-generation hybrid timeseries/analytics processing database in the cloud
</h3>
@@ -53,8 +58,9 @@ To compile GreptimeDB from source, you'll need:
install correct Rust version for you.
- Protobuf: `protoc` is required for compiling `.proto` files. `protobuf` is
available from major package manager on macos and linux distributions. You can
find an installation instructions
[here](https://grpc.io/docs/protoc-installation/).
find an installation instructions [here](https://grpc.io/docs/protoc-installation/).
**Note that `protoc` version needs to be >= 3.15** because we have used the `optional`
keyword. You can check it with `protoc --version`.
#### Build with Docker
@@ -114,7 +120,10 @@ about Kubernetes deployment, check our [docs](https://docs.greptime.com/).
4. Query the data:
```SQL
mysql> SELECT * FROM monitor;
SELECT * FROM monitor;
```
```TEXT
+-------+---------------------+------+--------+
| host | ts | cpu | memory |
+-------+---------------------+------+--------+
@@ -156,6 +165,8 @@ break things. Benchmark on development branch may not represent its potential
performance. We release pre-built binaries constantly for functional
evaluation. Do not use it in production at the moment.
For future plans, check out [GreptimeDB roadmap](https://github.com/GreptimeTeam/greptimedb/issues/669).
## Community
Our core team is thrilled too see you participate in any ways you like. When you are stuck, try to
@@ -169,7 +180,7 @@ community, please check out:
In addition, you may:
- View our official [Blog](https://greptime.com/blog)
- View our official [Blog](https://greptime.com/blogs/index)
- Connect us with [Linkedin](https://www.linkedin.com/company/greptime/)
- Follow us on [Twitter](https://twitter.com/greptime)
@@ -189,3 +200,4 @@ Please refer to [contribution guidelines](CONTRIBUTING.md) for more information.
- GreptimeDB's query engine is powered by [Apache Arrow DataFusion](https://github.com/apache/arrow-datafusion).
- [OpenDAL](https://github.com/datafuselabs/opendal) from [Datafuse Labs](https://github.com/datafuselabs) gives GreptimeDB a very general and elegant data access abstraction layer.
- GreptimeDBs meta service is based on [etcd](https://etcd.io/).
- GreptimeDB uses [RustPython](https://github.com/RustPython/RustPython) for experimental embedded python scripting.

View File

@@ -5,10 +5,10 @@ edition = "2021"
license = "Apache-2.0"
[dependencies]
arrow = "10"
arrow = "26.0.0"
clap = { version = "4.0", features = ["derive"] }
client = { path = "../src/client" }
indicatif = "0.17.1"
itertools = "0.10.5"
parquet = { version = "*" }
parquet = "26.0.0"
tokio = { version = "1.21", features = ["full"] }

View File

@@ -20,7 +20,6 @@
use std::collections::HashMap;
use std::path::{Path, PathBuf};
use std::sync::Arc;
use std::time::Instant;
use arrow::array::{ArrayRef, PrimitiveArray, StringArray, TimestampNanosecondArray};
@@ -28,14 +27,11 @@ use arrow::datatypes::{DataType, Float64Type, Int64Type};
use arrow::record_batch::RecordBatch;
use clap::Parser;
use client::admin::Admin;
use client::api::v1::codec::InsertBatch;
use client::api::v1::column::Values;
use client::api::v1::{insert_expr, Column, ColumnDataType, ColumnDef, CreateExpr, InsertExpr};
use client::api::v1::{Column, ColumnDataType, ColumnDef, CreateExpr, InsertExpr};
use client::{Client, Database, Select};
use indicatif::{MultiProgress, ProgressBar, ProgressStyle};
use parquet::arrow::{ArrowReader, ParquetFileArrowReader};
use parquet::file::reader::FileReader;
use parquet::file::serialized_reader::SerializedFileReader;
use parquet::arrow::arrow_reader::ParquetRecordBatchReaderBuilder;
use tokio::task::JoinSet;
const DATABASE_NAME: &str = "greptime";
@@ -87,10 +83,14 @@ async fn write_data(
pb_style: ProgressStyle,
) -> u128 {
let file = std::fs::File::open(&path).unwrap();
let file_reader = Arc::new(SerializedFileReader::new(file).unwrap());
let row_num = file_reader.metadata().file_metadata().num_rows();
let record_batch_reader = ParquetFileArrowReader::new(file_reader)
.get_record_reader(batch_size)
let record_batch_reader_builder = ParquetRecordBatchReaderBuilder::try_new(file).unwrap();
let row_num = record_batch_reader_builder
.metadata()
.file_metadata()
.num_rows();
let record_batch_reader = record_batch_reader_builder
.with_batch_size(batch_size)
.build()
.unwrap();
let progress_bar = mpb.add(ProgressBar::new(row_num as _));
progress_bar.set_style(pb_style);
@@ -100,16 +100,13 @@ async fn write_data(
for record_batch in record_batch_reader {
let record_batch = record_batch.unwrap();
let row_count = record_batch.num_rows();
let insert_batch = convert_record_batch(record_batch).into();
let (columns, row_count) = convert_record_batch(record_batch);
let insert_expr = InsertExpr {
schema_name: "public".to_string(),
table_name: TABLE_NAME.to_string(),
expr: Some(insert_expr::Expr::Values(insert_expr::Values {
values: vec![insert_batch],
})),
options: HashMap::default(),
region_number: 0,
columns,
row_count,
};
let now = Instant::now();
db.insert(insert_expr).await.unwrap();
@@ -125,7 +122,7 @@ async fn write_data(
total_rpc_elapsed_ms
}
fn convert_record_batch(record_batch: RecordBatch) -> InsertBatch {
fn convert_record_batch(record_batch: RecordBatch) -> (Vec<Column>, u32) {
let schema = record_batch.schema();
let fields = schema.fields();
let row_count = record_batch.num_rows();
@@ -143,10 +140,7 @@ fn convert_record_batch(record_batch: RecordBatch) -> InsertBatch {
columns.push(column);
}
InsertBatch {
columns,
row_count: row_count as _,
}
(columns, row_count as _)
}
fn build_values(column: &ArrayRef) -> Values {
@@ -217,9 +211,10 @@ fn build_values(column: &ArrayRef) -> Values {
| DataType::FixedSizeList(_, _)
| DataType::LargeList(_)
| DataType::Struct(_)
| DataType::Union(_, _)
| DataType::Union(_, _, _)
| DataType::Dictionary(_, _)
| DataType::Decimal(_, _)
| DataType::Decimal128(_, _)
| DataType::Decimal256(_, _)
| DataType::Map(_, _) => todo!(),
}
}

View File

@@ -7,3 +7,4 @@ coverage:
patch: off
ignore:
- "**/error*.rs" # ignore all error.rs files
- "tests/runner/*.rs" # ignore integration test runner

View File

@@ -3,15 +3,16 @@ mode = 'distributed'
rpc_addr = '127.0.0.1:3001'
wal_dir = '/tmp/greptimedb/wal'
rpc_runtime_size = 8
mysql_addr = '127.0.0.1:3306'
mysql_addr = '127.0.0.1:4406'
mysql_runtime_size = 4
enable_memory_catalog = false
[storage]
type = 'File'
data_dir = '/tmp/greptimedb/data/'
[meta_client_opts]
metasrv_addr = '1.1.1.1:3002'
metasrv_addrs = ['127.0.0.1:3002']
timeout_millis = 3000
connect_timeout_millis = 5000
tcp_nodelay = false

View File

@@ -1,9 +1,12 @@
mode = 'distributed'
datanode_rpc_addr = '127.0.0.1:3001'
http_addr = '127.0.0.1:4000'
[http_options]
addr = '127.0.0.1:4000'
timeout = "30s"
[meta_client_opts]
metasrv_addr = '1.1.1.1:3002'
metasrv_addrs = ['127.0.0.1:3002']
timeout_millis = 3000
connect_timeout_millis = 5000
tcp_nodelay = false

View File

@@ -1,9 +1,11 @@
node_id = 0
mode = 'standalone'
http_addr = '127.0.0.1:4000'
datanode_mysql_addr = '127.0.0.1:3306'
datanode_mysql_runtime_size = 4
wal_dir = '/tmp/greptimedb/wal/'
enable_memory_catalog = false
[http_options]
addr = '127.0.0.1:4000'
timeout = "30s"
[storage]
type = 'File'

View File

@@ -55,7 +55,7 @@ The DataFusion basically execute aggregate like this:
2. Call `update_batch` on each accumulator with partitioned data, to let you update your aggregate calculation.
3. Call `state` to get each accumulator's internal state, the medial calculation result.
4. Call `merge_batch` to merge all accumulator's internal state to one.
5. Execute `evalute` on the chosen one to get the final calculation result.
5. Execute `evaluate` on the chosen one to get the final calculation result.
Once you know the meaning of each method, you can easily write your accumulator. You can refer to `Median` accumulator or `SUM` accumulator defined in file `my_sum_udaf_example.rs` for more details.
@@ -63,7 +63,7 @@ Once you know the meaning of each method, you can easily write your accumulator.
You can call `register_aggregate_function` method in query engine to register your aggregate function. To do that, you have to new an instance of struct `AggregateFunctionMeta`. The struct has three fields, first is the name of your aggregate function's name. The function name is case-sensitive due to DataFusion's restriction. We strongly recommend using lowercase for your name. If you have to use uppercase name, wrap your aggregate function with quotation marks. For example, if you define an aggregate function named "my_aggr", you can use "`SELECT MY_AGGR(x)`"; if you define "my_AGGR", you have to use "`SELECT "my_AGGR"(x)`".
The second field is arg_counts ,the count of the arguments. Like accumulator `percentile`, caculating the p_number of the column. We need to input the value of column and the value of p to cacalate, and so the count of the arguments is two.
The second field is arg_counts ,the count of the arguments. Like accumulator `percentile`, calculating the p_number of the column. We need to input the value of column and the value of p to cacalate, and so the count of the arguments is two.
The third field is a function about how to create your accumulator creator that you defined in step 1 above. Create creator, that's a bit intertwined, but it is how we make DataFusion use a newly created aggregate function each time it executes a SQL, preventing the stored input types from affecting each other. The key detail can be starting looking at our `DfContextProviderAdapter` struct's `get_aggregate_meta` method.

Binary file not shown.

After

Width:  |  Height:  |  Size: 25 KiB

View File

@@ -7,8 +7,8 @@ license = "Apache-2.0"
[dependencies]
common-base = { path = "../common/base" }
common-time = { path = "../common/time" }
common-error = { path = "../common/error" }
common-time = { path = "../common/time" }
datatypes = { path = "../datatypes" }
prost = "0.11"
snafu = { version = "0.7", features = ["backtraces"] }

View File

@@ -20,9 +20,7 @@ fn main() {
.file_descriptor_set_path(default_out_dir.join("greptime_fd.bin"))
.compile(
&[
"greptime/v1/insert.proto",
"greptime/v1/select.proto",
"greptime/v1/physical_plan.proto",
"greptime/v1/greptime.proto",
"greptime/v1/meta/common.proto",
"greptime/v1/meta/heartbeat.proto",

View File

@@ -20,6 +20,7 @@ message AdminExpr {
CreateExpr create = 2;
AlterExpr alter = 3;
CreateDatabaseExpr create_database = 4;
DropTableExpr drop_table = 5;
}
}
@@ -51,18 +52,33 @@ message AlterExpr {
string table_name = 3;
oneof kind {
AddColumns add_columns = 4;
DropColumns drop_columns = 5;
}
}
message DropTableExpr {
string catalog_name = 1;
string schema_name = 2;
string table_name = 3;
}
message AddColumns {
repeated AddColumn add_columns = 1;
}
message DropColumns {
repeated DropColumn drop_columns = 1;
}
message AddColumn {
ColumnDef column_def = 1;
bool is_key = 2;
}
message DropColumn {
string name = 1;
}
message CreateDatabaseExpr {
//TODO(hl): maybe rename to schema_name?
string database_name = 1;

View File

@@ -32,7 +32,10 @@ message Column {
repeated int32 date_values = 14;
repeated int64 datetime_values = 15;
repeated int64 ts_millis_values = 16;
repeated int64 ts_second_values = 16;
repeated int64 ts_millisecond_values = 17;
repeated int64 ts_microsecond_values = 18;
repeated int64 ts_nanosecond_values = 19;
}
// The array of non-null values in this column.
//
@@ -75,5 +78,8 @@ enum ColumnDataType {
STRING = 12;
DATE = 13;
DATETIME = 14;
TIMESTAMP = 15;
TIMESTAMP_SECOND = 15;
TIMESTAMP_MILLISECOND = 16;
TIMESTAMP_MICROSECOND = 17;
TIMESTAMP_NANOSECOND = 18;
}

View File

@@ -2,6 +2,7 @@ syntax = "proto3";
package greptime.v1;
import "greptime/v1/column.proto";
import "greptime/v1/common.proto";
message DatabaseRequest {
@@ -28,39 +29,23 @@ message SelectExpr {
oneof expr {
string sql = 1;
bytes logical_plan = 2;
PhysicalPlan physical_plan = 15;
}
}
message PhysicalPlan {
bytes original_ql = 1;
bytes plan = 2;
}
message InsertExpr {
string schema_name = 1;
string table_name = 2;
message Values {
repeated bytes values = 1;
}
// Data is represented here.
repeated Column columns = 3;
oneof expr {
Values values = 3;
// The row_count of all columns, which include null and non-null values.
//
// Note: the row_count of all columns in a InsertExpr must be same.
uint32 row_count = 4;
// TODO(LFC): Remove field "sql" in InsertExpr.
// When Frontend instance received an insertion SQL (`insert into ...`), it's anticipated to parse the SQL and
// assemble the values to insert to feed Datanode. In other words, inserting data through Datanode instance's GRPC
// interface shouldn't use SQL directly.
// Then why the "sql" field exists here? It's because the Frontend needs table schema to create the values to insert,
// which is currently not able to find anywhere. (Maybe the table schema is suppose to be fetched from Meta?)
// The "sql" field is meant to be removed in the future.
string sql = 4;
}
/// The region number of current insert request.
// The region number of current insert request.
uint32 region_number = 5;
map<string, bytes> options = 6;
}
// TODO(jiachun)

View File

@@ -1,14 +0,0 @@
syntax = "proto3";
package greptime.v1.codec;
import "greptime/v1/column.proto";
message InsertBatch {
repeated Column columns = 1;
uint32 row_count = 2;
}
message RegionNumber {
uint32 id = 1;
}

View File

@@ -39,7 +39,7 @@ message NodeStat {
uint64 wcus = 2;
// Table number in this node
uint64 table_num = 3;
// Regon number in this node
// Region number in this node
uint64 region_num = 4;
double cpu_usage = 5;

View File

@@ -5,6 +5,8 @@ package greptime.v1.meta;
import "greptime/v1/meta/common.proto";
service Router {
rpc Create(CreateRequest) returns (RouteResponse) {}
// Fetch routing information for tables. The smallest unit is the complete
// routing information(all regions) of a table.
//
@@ -26,7 +28,14 @@ service Router {
//
rpc Route(RouteRequest) returns (RouteResponse) {}
rpc Create(CreateRequest) returns (RouteResponse) {}
rpc Delete(DeleteRequest) returns (RouteResponse) {}
}
message CreateRequest {
RequestHeader header = 1;
TableName table_name = 2;
repeated Partition partitions = 3;
}
message RouteRequest {
@@ -35,6 +44,12 @@ message RouteRequest {
repeated TableName table_names = 2;
}
message DeleteRequest {
RequestHeader header = 1;
TableName table_name = 2;
}
message RouteResponse {
ResponseHeader header = 1;
@@ -42,13 +57,6 @@ message RouteResponse {
repeated TableRoute table_routes = 3;
}
message CreateRequest {
RequestHeader header = 1;
TableName table_name = 2;
repeated Partition partitions = 3;
}
message TableRoute {
Table table = 1;
repeated RegionRoute region_routes = 2;

View File

@@ -20,6 +20,9 @@ service Store {
// DeleteRange deletes the given range from the key-value store.
rpc DeleteRange(DeleteRangeRequest) returns (DeleteRangeResponse);
// MoveValue atomically renames the key to the given updated key.
rpc MoveValue(MoveValueRequest) returns (MoveValueResponse);
}
message RangeRequest {
@@ -136,3 +139,21 @@ message DeleteRangeResponse {
// returned.
repeated KeyValue prev_kvs = 3;
}
message MoveValueRequest {
RequestHeader header = 1;
// If from_key dose not exist, return the value of to_key (if it exists).
// If from_key exists, move the value of from_key to to_key (i.e. rename),
// and return the value.
bytes from_key = 2;
bytes to_key = 3;
}
message MoveValueResponse {
ResponseHeader header = 1;
// If from_key dose not exist, return the value of to_key (if it exists).
// If from_key exists, return the value of from_key.
KeyValue kv = 2;
}

View File

@@ -1,33 +0,0 @@
syntax = "proto3";
package greptime.v1.codec;
message PhysicalPlanNode {
oneof PhysicalPlanType {
ProjectionExecNode projection = 1;
MockInputExecNode mock = 99;
// TODO(fys): impl other physical plan node
}
}
message ProjectionExecNode {
PhysicalPlanNode input = 1;
repeated PhysicalExprNode expr = 2;
repeated string expr_name = 3;
}
message PhysicalExprNode {
oneof ExprType {
PhysicalColumn column = 1;
// TODO(fys): impl other physical expr node
}
}
message PhysicalColumn {
string name = 1;
uint64 index = 2;
}
message MockInputExecNode {
string name = 1;
}

View File

@@ -33,6 +33,28 @@ pub enum Error {
from: ConcreteDataType,
backtrace: Backtrace,
},
#[snafu(display(
"Failed to convert column default constraint, column: {}, source: {}",
column,
source
))]
ConvertColumnDefaultConstraint {
column: String,
#[snafu(backtrace)]
source: datatypes::error::Error,
},
#[snafu(display(
"Invalid column default constraint, column: {}, source: {}",
column,
source
))]
InvalidColumnDefaultConstraint {
column: String,
#[snafu(backtrace)]
source: datatypes::error::Error,
},
}
impl ErrorExt for Error {
@@ -40,6 +62,8 @@ impl ErrorExt for Error {
match self {
Error::UnknownColumnDataType { .. } => StatusCode::InvalidArguments,
Error::IntoColumnDataType { .. } => StatusCode::Unexpected,
Error::ConvertColumnDefaultConstraint { source, .. }
| Error::InvalidColumnDefaultConstraint { source, .. } => source.status_code(),
}
}
fn backtrace_opt(&self) -> Option<&Backtrace> {

View File

@@ -15,6 +15,7 @@
use common_base::BitVec;
use common_time::timestamp::TimeUnit;
use datatypes::prelude::ConcreteDataType;
use datatypes::types::TimestampType;
use datatypes::value::Value;
use datatypes::vectors::VectorRef;
use snafu::prelude::*;
@@ -56,7 +57,16 @@ impl From<ColumnDataTypeWrapper> for ConcreteDataType {
ColumnDataType::String => ConcreteDataType::string_datatype(),
ColumnDataType::Date => ConcreteDataType::date_datatype(),
ColumnDataType::Datetime => ConcreteDataType::datetime_datatype(),
ColumnDataType::Timestamp => ConcreteDataType::timestamp_millis_datatype(),
ColumnDataType::TimestampSecond => ConcreteDataType::timestamp_second_datatype(),
ColumnDataType::TimestampMillisecond => {
ConcreteDataType::timestamp_millisecond_datatype()
}
ColumnDataType::TimestampMicrosecond => {
ConcreteDataType::timestamp_microsecond_datatype()
}
ColumnDataType::TimestampNanosecond => {
ConcreteDataType::timestamp_nanosecond_datatype()
}
}
}
}
@@ -81,7 +91,12 @@ impl TryFrom<ConcreteDataType> for ColumnDataTypeWrapper {
ConcreteDataType::String(_) => ColumnDataType::String,
ConcreteDataType::Date(_) => ColumnDataType::Date,
ConcreteDataType::DateTime(_) => ColumnDataType::Datetime,
ConcreteDataType::Timestamp(_) => ColumnDataType::Timestamp,
ConcreteDataType::Timestamp(unit) => match unit {
TimestampType::Second(_) => ColumnDataType::TimestampSecond,
TimestampType::Millisecond(_) => ColumnDataType::TimestampMillisecond,
TimestampType::Microsecond(_) => ColumnDataType::TimestampMicrosecond,
TimestampType::Nanosecond(_) => ColumnDataType::TimestampNanosecond,
},
ConcreteDataType::Null(_) | ConcreteDataType::List(_) => {
return error::IntoColumnDataTypeSnafu { from: datatype }.fail()
}
@@ -153,8 +168,20 @@ impl Values {
datetime_values: Vec::with_capacity(capacity),
..Default::default()
},
ColumnDataType::Timestamp => Values {
ts_millis_values: Vec::with_capacity(capacity),
ColumnDataType::TimestampSecond => Values {
ts_second_values: Vec::with_capacity(capacity),
..Default::default()
},
ColumnDataType::TimestampMillisecond => Values {
ts_millisecond_values: Vec::with_capacity(capacity),
..Default::default()
},
ColumnDataType::TimestampMicrosecond => Values {
ts_microsecond_values: Vec::with_capacity(capacity),
..Default::default()
},
ColumnDataType::TimestampNanosecond => Values {
ts_nanosecond_values: Vec::with_capacity(capacity),
..Default::default()
},
}
@@ -187,9 +214,12 @@ impl Column {
Value::Binary(val) => values.binary_values.push(val.to_vec()),
Value::Date(val) => values.date_values.push(val.val()),
Value::DateTime(val) => values.datetime_values.push(val.val()),
Value::Timestamp(val) => values
.ts_millis_values
.push(val.convert_to(TimeUnit::Millisecond)),
Value::Timestamp(val) => match val.unit() {
TimeUnit::Second => values.ts_second_values.push(val.value()),
TimeUnit::Millisecond => values.ts_millisecond_values.push(val.value()),
TimeUnit::Microsecond => values.ts_microsecond_values.push(val.value()),
TimeUnit::Nanosecond => values.ts_nanosecond_values.push(val.value()),
},
Value::List(_) => unreachable!(),
});
self.null_mask = null_mask.into_vec();
@@ -200,7 +230,10 @@ impl Column {
mod tests {
use std::sync::Arc;
use datatypes::vectors::BooleanVector;
use datatypes::vectors::{
BooleanVector, TimestampMicrosecondVector, TimestampMillisecondVector,
TimestampNanosecondVector, TimestampSecondVector,
};
use super::*;
@@ -258,8 +291,8 @@ mod tests {
let values = values.datetime_values;
assert_eq!(2, values.capacity());
let values = Values::with_capacity(ColumnDataType::Timestamp, 2);
let values = values.ts_millis_values;
let values = Values::with_capacity(ColumnDataType::TimestampMillisecond, 2);
let values = values.ts_millisecond_values;
assert_eq!(2, values.capacity());
}
@@ -326,8 +359,8 @@ mod tests {
ColumnDataTypeWrapper(ColumnDataType::Datetime).into()
);
assert_eq!(
ConcreteDataType::timestamp_millis_datatype(),
ColumnDataTypeWrapper(ColumnDataType::Timestamp).into()
ConcreteDataType::timestamp_millisecond_datatype(),
ColumnDataTypeWrapper(ColumnDataType::TimestampMillisecond).into()
);
}
@@ -394,8 +427,8 @@ mod tests {
ConcreteDataType::datetime_datatype().try_into().unwrap()
);
assert_eq!(
ColumnDataTypeWrapper(ColumnDataType::Timestamp),
ConcreteDataType::timestamp_millis_datatype()
ColumnDataTypeWrapper(ColumnDataType::TimestampMillisecond),
ConcreteDataType::timestamp_millisecond_datatype()
.try_into()
.unwrap()
);
@@ -412,7 +445,48 @@ mod tests {
assert!(result.is_err());
assert_eq!(
result.unwrap_err().to_string(),
"Failed to create column datatype from List(ListType { inner: Boolean(BooleanType) })"
"Failed to create column datatype from List(ListType { item_type: Boolean(BooleanType) })"
);
}
#[test]
fn test_column_put_timestamp_values() {
let mut column = Column {
column_name: "test".to_string(),
semantic_type: 0,
values: Some(Values {
..Default::default()
}),
null_mask: vec![],
datatype: 0,
};
let vector = Arc::new(TimestampNanosecondVector::from_vec(vec![1, 2, 3]));
column.push_vals(3, vector);
assert_eq!(
vec![1, 2, 3],
column.values.as_ref().unwrap().ts_nanosecond_values
);
let vector = Arc::new(TimestampMillisecondVector::from_vec(vec![4, 5, 6]));
column.push_vals(3, vector);
assert_eq!(
vec![4, 5, 6],
column.values.as_ref().unwrap().ts_millisecond_values
);
let vector = Arc::new(TimestampMicrosecondVector::from_vec(vec![7, 8, 9]));
column.push_vals(3, vector);
assert_eq!(
vec![7, 8, 9],
column.values.as_ref().unwrap().ts_microsecond_values
);
let vector = Arc::new(TimestampSecondVector::from_vec(vec![10, 11, 12]));
column.push_vals(3, vector);
assert_eq!(
vec![10, 11, 12],
column.values.as_ref().unwrap().ts_second_values
);
}

View File

@@ -15,7 +15,7 @@
pub use prost::DecodeError;
use prost::Message;
use crate::v1::codec::{InsertBatch, PhysicalPlanNode, RegionNumber, SelectResult};
use crate::v1::codec::SelectResult;
use crate::v1::meta::TableRouteValue;
macro_rules! impl_convert_with_bytes {
@@ -36,10 +36,7 @@ macro_rules! impl_convert_with_bytes {
};
}
impl_convert_with_bytes!(InsertBatch);
impl_convert_with_bytes!(SelectResult);
impl_convert_with_bytes!(PhysicalPlanNode);
impl_convert_with_bytes!(RegionNumber);
impl_convert_with_bytes!(TableRouteValue);
#[cfg(test)]
@@ -51,52 +48,6 @@ mod tests {
const SEMANTIC_TAG: i32 = 0;
#[test]
fn test_convert_insert_batch() {
let insert_batch = mock_insert_batch();
let bytes: Vec<u8> = insert_batch.into();
let insert: InsertBatch = bytes.deref().try_into().unwrap();
assert_eq!(8, insert.row_count);
assert_eq!(1, insert.columns.len());
let column = &insert.columns[0];
assert_eq!("foo", column.column_name);
assert_eq!(SEMANTIC_TAG, column.semantic_type);
assert_eq!(vec![1], column.null_mask);
assert_eq!(
vec![2, 3, 4, 5, 6, 7, 8],
column.values.as_ref().unwrap().i32_values
);
}
#[should_panic]
#[test]
fn test_convert_insert_batch_wrong() {
let insert_batch = mock_insert_batch();
let mut bytes: Vec<u8> = insert_batch.into();
// modify some bytes
bytes[0] = 0b1;
bytes[1] = 0b1;
let insert: InsertBatch = bytes.deref().try_into().unwrap();
assert_eq!(8, insert.row_count);
assert_eq!(1, insert.columns.len());
let column = &insert.columns[0];
assert_eq!("foo", column.column_name);
assert_eq!(SEMANTIC_TAG, column.semantic_type);
assert_eq!(vec![1], column.null_mask);
assert_eq!(
vec![2, 3, 4, 5, 6, 7, 8],
column.values.as_ref().unwrap().i32_values
);
}
#[test]
fn test_convert_select_result() {
let select_result = mock_select_result();
@@ -143,35 +94,6 @@ mod tests {
);
}
#[test]
fn test_convert_region_id() {
let region_id = RegionNumber { id: 12 };
let bytes: Vec<u8> = region_id.into();
let region_id: RegionNumber = bytes.deref().try_into().unwrap();
assert_eq!(12, region_id.id);
}
fn mock_insert_batch() -> InsertBatch {
let values = column::Values {
i32_values: vec![2, 3, 4, 5, 6, 7, 8],
..Default::default()
};
let null_mask = vec![1];
let column = Column {
column_name: "foo".to_string(),
semantic_type: SEMANTIC_TAG,
values: Some(values),
null_mask,
..Default::default()
};
InsertBatch {
columns: vec![column],
row_count: 8,
}
}
fn mock_select_result() -> SelectResult {
let values = column::Values {
i32_values: vec![2, 3, 4, 5, 6, 7, 8],

View File

@@ -21,4 +21,5 @@ pub mod codec {
tonic::include_proto!("greptime.v1.codec");
}
mod column_def;
pub mod meta;

View File

@@ -0,0 +1,38 @@
// Copyright 2022 Greptime Team
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
use datatypes::schema::{ColumnDefaultConstraint, ColumnSchema};
use snafu::ResultExt;
use crate::error::{self, Result};
use crate::helper::ColumnDataTypeWrapper;
use crate::v1::ColumnDef;
impl ColumnDef {
pub fn try_as_column_schema(&self) -> Result<ColumnSchema> {
let data_type = ColumnDataTypeWrapper::try_new(self.datatype)?;
let constraint = match &self.default_constraint {
None => None,
Some(v) => Some(
ColumnDefaultConstraint::try_from(&v[..])
.context(error::ConvertColumnDefaultConstraintSnafu { column: &self.name })?,
),
};
ColumnSchema::new(&self.name, data_type.into(), self.is_nullable)
.with_default_constraint(constraint)
.context(error::InvalidColumnDefaultConstraintSnafu { column: &self.name })
}
}

View File

@@ -145,10 +145,12 @@ gen_set_header!(HeartbeatRequest);
gen_set_header!(RouteRequest);
gen_set_header!(CreateRequest);
gen_set_header!(RangeRequest);
gen_set_header!(DeleteRequest);
gen_set_header!(PutRequest);
gen_set_header!(BatchPutRequest);
gen_set_header!(CompareAndPutRequest);
gen_set_header!(DeleteRangeRequest);
gen_set_header!(MoveValueRequest);
#[cfg(test)]
mod tests {

View File

@@ -19,15 +19,12 @@ common-recordbatch = { path = "../common/recordbatch" }
common-runtime = { path = "../common/runtime" }
common-telemetry = { path = "../common/telemetry" }
common-time = { path = "../common/time" }
datafusion = { git = "https://github.com/apache/arrow-datafusion.git", branch = "arrow2", features = [
"simd",
] }
datafusion = "14.0.0"
datatypes = { path = "../datatypes" }
futures = "0.3"
futures-util = "0.3"
lazy_static = "1.4"
meta-client = { path = "../meta-client" }
opendal = "0.17"
regex = "1.6"
serde = "1.0"
serde_json = "1.0"
@@ -39,9 +36,8 @@ tokio = { version = "1.18", features = ["full"] }
[dev-dependencies]
chrono = "0.4"
log-store = { path = "../log-store" }
mito = { path = "../mito", features = ["test"] }
object-store = { path = "../object-store" }
opendal = "0.17"
storage = { path = "../storage" }
mito = { path = "../mito" }
tempdir = "0.3"
tokio = { version = "1.0", features = ["full"] }

View File

@@ -17,7 +17,7 @@ use std::any::Any;
use common_error::ext::{BoxedError, ErrorExt};
use common_error::prelude::{Snafu, StatusCode};
use datafusion::error::DataFusionError;
use datatypes::arrow;
use datatypes::prelude::ConcreteDataType;
use datatypes::schema::RawSchema;
use snafu::{Backtrace, ErrorCompat};
@@ -51,14 +51,12 @@ pub enum Error {
SystemCatalog { msg: String, backtrace: Backtrace },
#[snafu(display(
"System catalog table type mismatch, expected: binary, found: {:?} source: {}",
"System catalog table type mismatch, expected: binary, found: {:?}",
data_type,
source
))]
SystemCatalogTypeMismatch {
data_type: arrow::datatypes::DataType,
#[snafu(backtrace)]
source: datatypes::error::Error,
data_type: ConcreteDataType,
backtrace: Backtrace,
},
#[snafu(display("Invalid system catalog entry type: {:?}", entry_type))]
@@ -94,7 +92,7 @@ pub enum Error {
backtrace: Backtrace,
},
#[snafu(display("Table {} already exists", table))]
#[snafu(display("Table `{}` already exists", table))]
TableExists { table: String, backtrace: Backtrace },
#[snafu(display("Schema {} already exists", schema))]
@@ -109,6 +107,12 @@ pub enum Error {
source: BoxedError,
},
#[snafu(display("Operation {} not implemented yet", operation))]
Unimplemented {
operation: String,
backtrace: Backtrace,
},
#[snafu(display("Failed to open table, table info: {}, source: {}", table_info, source))]
OpenTable {
table_info: String,
@@ -185,8 +189,8 @@ pub enum Error {
source: meta_client::error::Error,
},
#[snafu(display("Invalid table schema in catalog, source: {:?}", source))]
InvalidSchemaInCatalog {
#[snafu(display("Invalid table info in catalog, source: {}", source))]
InvalidTableInfoInCatalog {
#[snafu(backtrace)]
source: datatypes::error::Error,
},
@@ -216,11 +220,13 @@ impl ErrorExt for Error {
| Error::ValueDeserialize { .. }
| Error::Io { .. } => StatusCode::StorageUnavailable,
Error::RegisterTable { .. } | Error::SystemCatalogTypeMismatch { .. } => {
StatusCode::Internal
}
Error::ReadSystemCatalog { source, .. } => source.status_code(),
Error::SystemCatalogTypeMismatch { source, .. } => source.status_code(),
Error::InvalidCatalogValue { source, .. } => source.status_code(),
Error::RegisterTable { .. } => StatusCode::Internal,
Error::TableExists { .. } => StatusCode::TableAlreadyExists,
Error::SchemaExists { .. } => StatusCode::InvalidArguments,
@@ -233,8 +239,10 @@ impl ErrorExt for Error {
Error::SystemCatalogTableScan { source } => source.status_code(),
Error::SystemCatalogTableScanExec { source } => source.status_code(),
Error::InvalidTableSchema { source, .. } => source.status_code(),
Error::InvalidSchemaInCatalog { .. } => StatusCode::Unexpected,
Error::InvalidTableInfoInCatalog { .. } => StatusCode::Unexpected,
Error::Internal { source, .. } => source.status_code(),
Error::Unimplemented { .. } => StatusCode::Unsupported,
}
}
@@ -256,7 +264,6 @@ impl From<Error> for DataFusionError {
#[cfg(test)]
mod tests {
use common_error::mock::MockError;
use datatypes::arrow::datatypes::DataType;
use snafu::GenerateImplicitData;
use super::*;
@@ -305,11 +312,8 @@ mod tests {
assert_eq!(
StatusCode::Internal,
Error::SystemCatalogTypeMismatch {
data_type: DataType::Boolean,
source: datatypes::error::Error::UnsupportedArrowType {
arrow_type: DataType::Boolean,
backtrace: Backtrace::generate()
}
data_type: ConcreteDataType::binary_datatype(),
backtrace: Backtrace::generate(),
}
.status_code()
);

View File

@@ -15,44 +15,56 @@
use std::collections::HashMap;
use std::fmt::{Display, Formatter};
use common_catalog::error::{
DeserializeCatalogEntryValueSnafu, Error, InvalidCatalogSnafu, SerializeCatalogEntryValueSnafu,
};
use lazy_static::lazy_static;
use regex::Regex;
use serde::{Deserialize, Serialize, Serializer};
use snafu::{ensure, OptionExt, ResultExt};
use table::metadata::{RawTableMeta, TableId, TableVersion};
use table::metadata::{RawTableInfo, TableId, TableVersion};
use crate::consts::{
CATALOG_KEY_PREFIX, SCHEMA_KEY_PREFIX, TABLE_GLOBAL_KEY_PREFIX, TABLE_REGIONAL_KEY_PREFIX,
};
use crate::error::{
DeserializeCatalogEntryValueSnafu, Error, InvalidCatalogSnafu, SerializeCatalogEntryValueSnafu,
};
const CATALOG_KEY_PREFIX: &str = "__c";
const SCHEMA_KEY_PREFIX: &str = "__s";
const TABLE_GLOBAL_KEY_PREFIX: &str = "__tg";
const TABLE_REGIONAL_KEY_PREFIX: &str = "__tr";
const ALPHANUMERICS_NAME_PATTERN: &str = "[a-zA-Z_][a-zA-Z0-9_]*";
lazy_static! {
static ref CATALOG_KEY_PATTERN: Regex =
Regex::new(&format!("^{}-([a-zA-Z_]+)$", CATALOG_KEY_PREFIX)).unwrap();
static ref CATALOG_KEY_PATTERN: Regex = Regex::new(&format!(
"^{}-({})$",
CATALOG_KEY_PREFIX, ALPHANUMERICS_NAME_PATTERN
))
.unwrap();
}
lazy_static! {
static ref SCHEMA_KEY_PATTERN: Regex = Regex::new(&format!(
"^{}-([a-zA-Z_]+)-([a-zA-Z_]+)$",
SCHEMA_KEY_PREFIX
"^{}-({})-({})$",
SCHEMA_KEY_PREFIX, ALPHANUMERICS_NAME_PATTERN, ALPHANUMERICS_NAME_PATTERN
))
.unwrap();
}
lazy_static! {
static ref TABLE_GLOBAL_KEY_PATTERN: Regex = Regex::new(&format!(
"^{}-([a-zA-Z_]+)-([a-zA-Z_]+)-([a-zA-Z0-9_]+)$",
TABLE_GLOBAL_KEY_PREFIX
"^{}-({})-({})-({})$",
TABLE_GLOBAL_KEY_PREFIX,
ALPHANUMERICS_NAME_PATTERN,
ALPHANUMERICS_NAME_PATTERN,
ALPHANUMERICS_NAME_PATTERN
))
.unwrap();
}
lazy_static! {
static ref TABLE_REGIONAL_KEY_PATTERN: Regex = Regex::new(&format!(
"^{}-([a-zA-Z_]+)-([a-zA-Z_]+)-([a-zA-Z0-9_]+)-([0-9]+)$",
TABLE_REGIONAL_KEY_PREFIX
"^{}-({})-({})-({})-([0-9]+)$",
TABLE_REGIONAL_KEY_PREFIX,
ALPHANUMERICS_NAME_PATTERN,
ALPHANUMERICS_NAME_PATTERN,
ALPHANUMERICS_NAME_PATTERN
))
.unwrap();
}
@@ -126,17 +138,20 @@ impl TableGlobalKey {
/// Table global info contains necessary info for a datanode to create table regions, including
/// table id, table meta(schema...), region id allocation across datanodes.
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
pub struct TableGlobalValue {
/// Table id is the same across all datanodes.
pub id: TableId,
/// Id of datanode that created the global table info kv. only for debugging.
pub node_id: u64,
// TODO(LFC): Maybe remove it?
/// Allocation of region ids across all datanodes.
pub regions_id_map: HashMap<u64, Vec<u32>>,
// TODO(LFC): Too much for assembling the table schema that DistTable needs, find another way.
pub meta: RawTableMeta,
pub table_info: RawTableInfo,
}
impl TableGlobalValue {
pub fn table_id(&self) -> TableId {
self.table_info.ident.table_id
}
}
/// Table regional info that varies between datanode, so it contains a `node_id` field.
@@ -258,6 +273,10 @@ macro_rules! define_catalog_value {
.context(DeserializeCatalogEntryValueSnafu { raw: s.as_ref() })
}
pub fn from_bytes(bytes: impl AsRef<[u8]>) -> Result<Self, Error> {
Self::parse(&String::from_utf8_lossy(bytes.as_ref()))
}
pub fn as_bytes(&self) -> Result<Vec<u8>, Error> {
Ok(serde_json::to_string(self)
.context(SerializeCatalogEntryValueSnafu)?
@@ -279,6 +298,7 @@ define_catalog_value!(
mod tests {
use datatypes::prelude::ConcreteDataType;
use datatypes::schema::{ColumnSchema, RawSchema, Schema};
use table::metadata::{RawTableMeta, TableIdent, TableType};
use super::*;
@@ -339,11 +359,23 @@ mod tests {
region_numbers: vec![1],
};
let table_info = RawTableInfo {
ident: TableIdent {
table_id: 42,
version: 1,
},
name: "table_1".to_string(),
desc: Some("blah".to_string()),
catalog_name: "catalog_1".to_string(),
schema_name: "schema_1".to_string(),
meta,
table_type: TableType::Base,
};
let value = TableGlobalValue {
id: 42,
node_id: 0,
regions_id_map: HashMap::from([(0, vec![1, 2, 3])]),
meta,
table_info,
};
let serialized = serde_json::to_string(&value).unwrap();
let deserialized = TableGlobalValue::parse(&serialized).unwrap();

View File

@@ -15,6 +15,7 @@
#![feature(assert_matches)]
use std::any::Any;
use std::fmt::{Debug, Formatter};
use std::sync::Arc;
use common_telemetry::info;
@@ -28,6 +29,7 @@ use crate::error::{CreateTableSnafu, Result};
pub use crate::schema::{SchemaProvider, SchemaProviderRef};
pub mod error;
pub mod helper;
pub mod local;
pub mod remote;
pub mod schema;
@@ -83,12 +85,17 @@ pub trait CatalogManager: CatalogList {
/// Starts a catalog manager.
async fn start(&self) -> Result<()>;
/// Registers a table given given catalog/schema to catalog manager,
/// returns table registered.
async fn register_table(&self, request: RegisterTableRequest) -> Result<usize>;
/// Registers a table within given catalog/schema to catalog manager,
/// returns whether the table registered.
async fn register_table(&self, request: RegisterTableRequest) -> Result<bool>;
/// Register a schema with catalog name and schema name.
async fn register_schema(&self, request: RegisterSchemaRequest) -> Result<usize>;
/// Deregisters a table within given catalog/schema to catalog manager,
/// returns whether the table deregistered.
async fn deregister_table(&self, request: DeregisterTableRequest) -> Result<bool>;
/// Register a schema with catalog name and schema name. Retuens whether the
/// schema registered.
async fn register_schema(&self, request: RegisterSchemaRequest) -> Result<bool>;
/// Register a system table, should be called before starting the manager.
async fn register_system_table(&self, request: RegisterSystemTableRequest)
@@ -123,6 +130,25 @@ pub struct RegisterTableRequest {
pub table: TableRef,
}
impl Debug for RegisterTableRequest {
fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
f.debug_struct("RegisterTableRequest")
.field("catalog", &self.catalog)
.field("schema", &self.schema)
.field("table_name", &self.table_name)
.field("table_id", &self.table_id)
.field("table", &self.table.table_info())
.finish()
}
}
#[derive(Clone)]
pub struct DeregisterTableRequest {
pub catalog: String,
pub schema: String,
pub table_name: String,
}
#[derive(Debug, Clone)]
pub struct RegisterSchemaRequest {
pub catalog: String,

View File

@@ -21,7 +21,7 @@ use common_catalog::consts::{
SYSTEM_CATALOG_NAME, SYSTEM_CATALOG_TABLE_NAME,
};
use common_recordbatch::{RecordBatch, SendableRecordBatchStream};
use common_telemetry::info;
use common_telemetry::{error, info};
use datatypes::prelude::ScalarVector;
use datatypes::vectors::{BinaryVector, UInt8Vector};
use futures_util::lock::Mutex;
@@ -36,7 +36,7 @@ use table::TableRef;
use crate::error::{
CatalogNotFoundSnafu, IllegalManagerStateSnafu, OpenTableSnafu, ReadSystemCatalogSnafu, Result,
SchemaExistsSnafu, SchemaNotFoundSnafu, SystemCatalogSnafu, SystemCatalogTypeMismatchSnafu,
TableExistsSnafu, TableNotFoundSnafu,
TableExistsSnafu, TableNotFoundSnafu, UnimplementedSnafu,
};
use crate::local::memory::{MemoryCatalogManager, MemoryCatalogProvider, MemorySchemaProvider};
use crate::system::{
@@ -46,8 +46,8 @@ use crate::system::{
use crate::tables::SystemCatalog;
use crate::{
format_full_table_name, handle_system_table_request, CatalogList, CatalogManager,
CatalogProvider, CatalogProviderRef, RegisterSchemaRequest, RegisterSystemTableRequest,
RegisterTableRequest, SchemaProvider, SchemaProviderRef,
CatalogProvider, CatalogProviderRef, DeregisterTableRequest, RegisterSchemaRequest,
RegisterSystemTableRequest, RegisterTableRequest, SchemaProvider, SchemaProviderRef,
};
/// A `CatalogManager` consists of a system catalog and a bunch of user catalogs.
@@ -57,6 +57,7 @@ pub struct LocalCatalogManager {
engine: TableEngineRef,
next_table_id: AtomicU32,
init_lock: Mutex<bool>,
register_lock: Mutex<()>,
system_table_requests: Mutex<Vec<RegisterSystemTableRequest>>,
}
@@ -76,6 +77,7 @@ impl LocalCatalogManager {
engine,
next_table_id: AtomicU32::new(MIN_USER_TABLE_ID),
init_lock: Mutex::new(false),
register_lock: Mutex::new(()),
system_table_requests: Mutex::new(Vec::default()),
})
}
@@ -143,27 +145,34 @@ impl LocalCatalogManager {
/// Convert `RecordBatch` to a vector of `Entry`.
fn record_batch_to_entry(rb: RecordBatch) -> Result<Vec<Entry>> {
ensure!(
rb.df_recordbatch.columns().len() >= 6,
rb.num_columns() >= 6,
SystemCatalogSnafu {
msg: format!("Length mismatch: {}", rb.df_recordbatch.columns().len())
msg: format!("Length mismatch: {}", rb.num_columns())
}
);
let entry_type = UInt8Vector::try_from_arrow_array(&rb.df_recordbatch.columns()[0])
.with_context(|_| SystemCatalogTypeMismatchSnafu {
data_type: rb.df_recordbatch.columns()[ENTRY_TYPE_INDEX]
.data_type()
.clone(),
let entry_type = rb
.column(ENTRY_TYPE_INDEX)
.as_any()
.downcast_ref::<UInt8Vector>()
.with_context(|| SystemCatalogTypeMismatchSnafu {
data_type: rb.column(ENTRY_TYPE_INDEX).data_type(),
})?;
let key = BinaryVector::try_from_arrow_array(&rb.df_recordbatch.columns()[1])
.with_context(|_| SystemCatalogTypeMismatchSnafu {
data_type: rb.df_recordbatch.columns()[KEY_INDEX].data_type().clone(),
let key = rb
.column(KEY_INDEX)
.as_any()
.downcast_ref::<BinaryVector>()
.with_context(|| SystemCatalogTypeMismatchSnafu {
data_type: rb.column(KEY_INDEX).data_type(),
})?;
let value = BinaryVector::try_from_arrow_array(&rb.df_recordbatch.columns()[3])
.with_context(|_| SystemCatalogTypeMismatchSnafu {
data_type: rb.df_recordbatch.columns()[VALUE_INDEX].data_type().clone(),
let value = rb
.column(VALUE_INDEX)
.as_any()
.downcast_ref::<BinaryVector>()
.with_context(|| SystemCatalogTypeMismatchSnafu {
data_type: rb.column(VALUE_INDEX).data_type(),
})?;
let mut res = Vec::with_capacity(rb.num_rows());
@@ -241,6 +250,7 @@ impl LocalCatalogManager {
schema_name: t.schema_name.clone(),
table_name: t.table_name.clone(),
table_id: t.table_id,
region_numbers: vec![0],
};
let option = self
@@ -308,7 +318,7 @@ impl CatalogManager for LocalCatalogManager {
self.init().await
}
async fn register_table(&self, request: RegisterTableRequest) -> Result<usize> {
async fn register_table(&self, request: RegisterTableRequest) -> Result<bool> {
let started = self.init_lock.lock().await;
ensure!(
@@ -331,27 +341,50 @@ impl CatalogManager for LocalCatalogManager {
schema_info: format!("{}.{}", catalog_name, schema_name),
})?;
if schema.table_exist(&request.table_name)? {
return TableExistsSnafu {
table: format_full_table_name(catalog_name, schema_name, &request.table_name),
{
let _lock = self.register_lock.lock().await;
if let Some(existing) = schema.table(&request.table_name)? {
if existing.table_info().ident.table_id != request.table_id {
error!(
"Unexpected table register request: {:?}, existing: {:?}",
request,
existing.table_info()
);
return TableExistsSnafu {
table: format_full_table_name(
catalog_name,
schema_name,
&request.table_name,
),
}
.fail();
}
// Try to register table with same table id, just ignore.
Ok(false)
} else {
// table does not exist
self.system
.register_table(
catalog_name.clone(),
schema_name.clone(),
request.table_name.clone(),
request.table_id,
)
.await?;
schema.register_table(request.table_name, request.table)?;
Ok(true)
}
.fail();
}
self.system
.register_table(
catalog_name.clone(),
schema_name.clone(),
request.table_name.clone(),
request.table_id,
)
.await?;
schema.register_table(request.table_name, request.table)?;
Ok(1)
}
async fn register_schema(&self, request: RegisterSchemaRequest) -> Result<usize> {
async fn deregister_table(&self, _request: DeregisterTableRequest) -> Result<bool> {
UnimplementedSnafu {
operation: "deregister table",
}
.fail()
}
async fn register_schema(&self, request: RegisterSchemaRequest) -> Result<bool> {
let started = self.init_lock.lock().await;
ensure!(
*started,
@@ -366,17 +399,21 @@ impl CatalogManager for LocalCatalogManager {
.catalogs
.catalog(catalog_name)?
.context(CatalogNotFoundSnafu { catalog_name })?;
if catalog.schema(schema_name)?.is_some() {
return SchemaExistsSnafu {
schema: schema_name,
}
.fail();
{
let _lock = self.register_lock.lock().await;
ensure!(
catalog.schema(schema_name)?.is_none(),
SchemaExistsSnafu {
schema: schema_name,
}
);
self.system
.register_schema(request.catalog, schema_name.clone())
.await?;
catalog.register_schema(request.schema, Arc::new(MemorySchemaProvider::new()))?;
Ok(true)
}
self.system
.register_schema(request.catalog, schema_name.clone())
.await?;
catalog.register_schema(request.schema, Arc::new(MemorySchemaProvider::new()))?;
Ok(1)
}
async fn register_system_table(&self, request: RegisterSystemTableRequest) -> Result<()> {

View File

@@ -19,6 +19,7 @@ use std::sync::atomic::{AtomicU32, Ordering};
use std::sync::{Arc, RwLock};
use common_catalog::consts::MIN_USER_TABLE_ID;
use common_telemetry::error;
use snafu::OptionExt;
use table::metadata::TableId;
use table::table::TableIdProvider;
@@ -27,8 +28,8 @@ use table::TableRef;
use crate::error::{CatalogNotFoundSnafu, Result, SchemaNotFoundSnafu, TableExistsSnafu};
use crate::schema::SchemaProvider;
use crate::{
CatalogList, CatalogManager, CatalogProvider, CatalogProviderRef, RegisterSchemaRequest,
RegisterSystemTableRequest, RegisterTableRequest, SchemaProviderRef,
CatalogList, CatalogManager, CatalogProvider, CatalogProviderRef, DeregisterTableRequest,
RegisterSchemaRequest, RegisterSystemTableRequest, RegisterTableRequest, SchemaProviderRef,
};
/// Simple in-memory list of catalogs
@@ -69,7 +70,7 @@ impl CatalogManager for MemoryCatalogManager {
Ok(())
}
async fn register_table(&self, request: RegisterTableRequest) -> Result<usize> {
async fn register_table(&self, request: RegisterTableRequest) -> Result<bool> {
let catalogs = self.catalogs.write().unwrap();
let catalog = catalogs
.get(&request.catalog)
@@ -84,10 +85,28 @@ impl CatalogManager for MemoryCatalogManager {
})?;
schema
.register_table(request.table_name, request.table)
.map(|v| if v.is_some() { 0 } else { 1 })
.map(|v| v.is_none())
}
async fn register_schema(&self, request: RegisterSchemaRequest) -> Result<usize> {
async fn deregister_table(&self, request: DeregisterTableRequest) -> Result<bool> {
let catalogs = self.catalogs.write().unwrap();
let catalog = catalogs
.get(&request.catalog)
.context(CatalogNotFoundSnafu {
catalog_name: &request.catalog,
})?
.clone();
let schema = catalog
.schema(&request.schema)?
.with_context(|| SchemaNotFoundSnafu {
schema_info: format!("{}.{}", &request.catalog, &request.schema),
})?;
schema
.deregister_table(&request.table_name)
.map(|v| v.is_some())
}
async fn register_schema(&self, request: RegisterSchemaRequest) -> Result<bool> {
let catalogs = self.catalogs.write().unwrap();
let catalog = catalogs
.get(&request.catalog)
@@ -95,11 +114,12 @@ impl CatalogManager for MemoryCatalogManager {
catalog_name: &request.catalog,
})?;
catalog.register_schema(request.schema, Arc::new(MemorySchemaProvider::new()))?;
Ok(1)
Ok(true)
}
async fn register_system_table(&self, _request: RegisterSystemTableRequest) -> Result<()> {
unimplemented!()
// TODO(ruihang): support register system table request
Ok(())
}
fn schema(&self, catalog: &str, schema: &str) -> Result<Option<SchemaProviderRef>> {
@@ -251,11 +271,21 @@ impl SchemaProvider for MemorySchemaProvider {
}
fn register_table(&self, name: String, table: TableRef) -> Result<Option<TableRef>> {
if self.table_exist(name.as_str())? {
return TableExistsSnafu { table: name }.fail()?;
}
let mut tables = self.tables.write().unwrap();
Ok(tables.insert(name, table))
if let Some(existing) = tables.get(name.as_str()) {
// if table with the same name but different table id exists, then it's a fatal bug
if existing.table_info().ident.table_id != table.table_info().ident.table_id {
error!(
"Unexpected table register: {:?}, existing: {:?}",
table.table_info(),
existing.table_info()
);
return TableExistsSnafu { table: name }.fail()?;
}
Ok(Some(existing.clone()))
} else {
Ok(tables.insert(name, table))
}
}
fn deregister_table(&self, name: &str) -> Result<Option<TableRef>> {
@@ -315,7 +345,7 @@ mod tests {
.unwrap()
.is_none());
assert!(provider.table_exist(table_name).unwrap());
let other_table = NumbersTable::default();
let other_table = NumbersTable::new(12);
let result = provider.register_table(table_name.to_string(), Arc::new(other_table));
let err = result.err().unwrap();
assert!(err.backtrace_opt().is_some());
@@ -340,4 +370,34 @@ mod tests {
.downcast_ref::<MemoryCatalogManager>()
.unwrap();
}
#[tokio::test]
pub async fn test_catalog_deregister_table() {
let catalog = MemoryCatalogManager::default();
let schema = catalog
.schema(DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME)
.unwrap()
.unwrap();
let register_table_req = RegisterTableRequest {
catalog: DEFAULT_CATALOG_NAME.to_string(),
schema: DEFAULT_SCHEMA_NAME.to_string(),
table_name: "numbers".to_string(),
table_id: 2333,
table: Arc::new(NumbersTable::default()),
};
catalog.register_table(register_table_req).await.unwrap();
assert!(schema.table_exist("numbers").unwrap());
let deregister_table_req = DeregisterTableRequest {
catalog: DEFAULT_CATALOG_NAME.to_string(),
schema: DEFAULT_SCHEMA_NAME.to_string(),
table_name: "numbers".to_string(),
};
catalog
.deregister_table(deregister_table_req)
.await
.unwrap();
assert!(!schema.table_exist("numbers").unwrap());
}
}

View File

@@ -20,10 +20,6 @@ use std::sync::Arc;
use arc_swap::ArcSwap;
use async_stream::stream;
use common_catalog::consts::{DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME, MIN_USER_TABLE_ID};
use common_catalog::{
build_catalog_prefix, build_schema_prefix, build_table_global_prefix, CatalogKey, CatalogValue,
SchemaKey, SchemaValue, TableGlobalKey, TableGlobalValue, TableRegionalKey, TableRegionalValue,
};
use common_telemetry::{debug, info};
use futures::Stream;
use futures_util::StreamExt;
@@ -37,13 +33,17 @@ use tokio::sync::Mutex;
use crate::error::{
CatalogNotFoundSnafu, CreateTableSnafu, InvalidCatalogValueSnafu, InvalidTableSchemaSnafu,
OpenTableSnafu, Result, SchemaNotFoundSnafu, TableExistsSnafu,
OpenTableSnafu, Result, SchemaNotFoundSnafu, TableExistsSnafu, UnimplementedSnafu,
};
use crate::helper::{
build_catalog_prefix, build_schema_prefix, build_table_global_prefix, CatalogKey, CatalogValue,
SchemaKey, SchemaValue, TableGlobalKey, TableGlobalValue, TableRegionalKey, TableRegionalValue,
};
use crate::remote::{Kv, KvBackendRef};
use crate::{
handle_system_table_request, CatalogList, CatalogManager, CatalogProvider, CatalogProviderRef,
RegisterSchemaRequest, RegisterSystemTableRequest, RegisterTableRequest, SchemaProvider,
SchemaProviderRef,
DeregisterTableRequest, RegisterSchemaRequest, RegisterSystemTableRequest,
RegisterTableRequest, SchemaProvider, SchemaProviderRef,
};
/// Catalog manager based on metasrv.
@@ -154,8 +154,8 @@ impl RemoteCatalogManager {
}
let table_key = TableGlobalKey::parse(&String::from_utf8_lossy(&k))
.context(InvalidCatalogValueSnafu)?;
let table_value = TableGlobalValue::parse(&String::from_utf8_lossy(&v))
.context(InvalidCatalogValueSnafu)?;
let table_value =
TableGlobalValue::from_bytes(&v).context(InvalidCatalogValueSnafu)?;
info!(
"Found catalog table entry, key: {}, value: {:?}",
@@ -250,10 +250,7 @@ impl RemoteCatalogManager {
let table_ref = self.open_or_create_table(&table_key, &table_value).await?;
schema.register_table(table_key.table_name.to_string(), table_ref)?;
info!("Registered table {}", &table_key.table_name);
if table_value.id > max_table_id {
info!("Max table id: {} -> {}", max_table_id, table_value.id);
max_table_id = table_value.id;
}
max_table_id = max_table_id.max(table_value.table_id());
table_num += 1;
}
info!(
@@ -311,25 +308,33 @@ impl RemoteCatalogManager {
..
} = table_key;
let table_id = table_value.table_id();
let TableGlobalValue {
id,
meta,
table_info,
regions_id_map,
..
} = table_value;
// unwrap safety: checked in yielding this table when `iter_remote_tables`
let region_numbers = regions_id_map.get(&self.node_id).unwrap();
let request = OpenTableRequest {
catalog_name: catalog_name.clone(),
schema_name: schema_name.clone(),
table_name: table_name.clone(),
table_id: *id,
table_id,
region_numbers: region_numbers.clone(),
};
match self
.engine
.open_table(&context, request)
.await
.with_context(|_| OpenTableSnafu {
table_info: format!("{}.{}.{}, id:{}", catalog_name, schema_name, table_name, id,),
table_info: format!(
"{}.{}.{}, id:{}",
catalog_name, schema_name, table_name, table_id
),
})? {
Some(table) => {
info!(
@@ -344,6 +349,7 @@ impl RemoteCatalogManager {
catalog_name, schema_name, table_name
);
let meta = &table_info.meta;
let schema = meta
.schema
.clone()
@@ -353,13 +359,13 @@ impl RemoteCatalogManager {
schema: meta.schema.clone(),
})?;
let req = CreateTableRequest {
id: *id,
id: table_id,
catalog_name: catalog_name.clone(),
schema_name: schema_name.clone(),
table_name: table_name.clone(),
desc: None,
schema: Arc::new(schema),
region_numbers: regions_id_map.get(&self.node_id).unwrap().clone(), // this unwrap is safe because region_id_map is checked in `iter_remote_tables`
region_numbers: region_numbers.clone(),
primary_key_indices: meta.primary_key_indices.clone(),
create_if_not_exists: true,
table_options: meta.options.clone(),
@@ -371,7 +377,7 @@ impl RemoteCatalogManager {
.context(CreateTableSnafu {
table_info: format!(
"{}.{}.{}, id:{}",
&catalog_name, &schema_name, &table_name, id
&catalog_name, &schema_name, &table_name, table_id
),
})
}
@@ -405,7 +411,7 @@ impl CatalogManager for RemoteCatalogManager {
Ok(())
}
async fn register_table(&self, request: RegisterTableRequest) -> Result<usize> {
async fn register_table(&self, request: RegisterTableRequest) -> Result<bool> {
let catalog_name = request.catalog;
let schema_name = request.schema;
let catalog_provider = self.catalog(&catalog_name)?.context(CatalogNotFoundSnafu {
@@ -424,10 +430,17 @@ impl CatalogManager for RemoteCatalogManager {
.fail();
}
schema_provider.register_table(request.table_name, request.table)?;
Ok(1)
Ok(true)
}
async fn register_schema(&self, request: RegisterSchemaRequest) -> Result<usize> {
async fn deregister_table(&self, _request: DeregisterTableRequest) -> Result<bool> {
UnimplementedSnafu {
operation: "deregister table",
}
.fail()
}
async fn register_schema(&self, request: RegisterSchemaRequest) -> Result<bool> {
let catalog_name = request.catalog;
let schema_name = request.schema;
let catalog_provider = self.catalog(&catalog_name)?.context(CatalogNotFoundSnafu {
@@ -435,7 +448,7 @@ impl CatalogManager for RemoteCatalogManager {
})?;
let schema_provider = self.new_schema_provider(&catalog_name, &schema_name);
catalog_provider.register_schema(schema_name, schema_provider)?;
Ok(1)
Ok(true)
}
async fn register_system_table(&self, request: RegisterSystemTableRequest) -> Result<()> {

View File

@@ -21,14 +21,13 @@ use common_catalog::consts::{
SYSTEM_CATALOG_TABLE_ID, SYSTEM_CATALOG_TABLE_NAME,
};
use common_query::logical_plan::Expr;
use common_query::physical_plan::{PhysicalPlanRef, RuntimeEnv};
use common_query::physical_plan::{PhysicalPlanRef, SessionContext};
use common_recordbatch::SendableRecordBatchStream;
use common_telemetry::debug;
use common_time::timestamp::Timestamp;
use common_time::util;
use datatypes::prelude::{ConcreteDataType, ScalarVector};
use datatypes::schema::{ColumnSchema, Schema, SchemaBuilder, SchemaRef};
use datatypes::vectors::{BinaryVector, TimestampVector, UInt8Vector};
use datatypes::vectors::{BinaryVector, TimestampMillisecondVector, UInt8Vector};
use serde::{Deserialize, Serialize};
use snafu::{ensure, OptionExt, ResultExt};
use table::engine::{EngineContext, TableEngineRef};
@@ -43,7 +42,6 @@ use crate::error::{
pub const ENTRY_TYPE_INDEX: usize = 0;
pub const KEY_INDEX: usize = 1;
pub const TIMESTAMP_INDEX: usize = 2;
pub const VALUE_INDEX: usize = 3;
pub struct SystemCatalogTable {
@@ -87,6 +85,7 @@ impl SystemCatalogTable {
schema_name: INFORMATION_SCHEMA_NAME.to_string(),
table_name: SYSTEM_CATALOG_TABLE_NAME.to_string(),
table_id: SYSTEM_CATALOG_TABLE_ID,
region_numbers: vec![0],
};
let schema = Arc::new(build_system_catalog_schema());
let ctx = EngineContext::default();
@@ -110,7 +109,7 @@ impl SystemCatalogTable {
desc: Some("System catalog table".to_string()),
schema: schema.clone(),
region_numbers: vec![0],
primary_key_indices: vec![ENTRY_TYPE_INDEX, KEY_INDEX, TIMESTAMP_INDEX],
primary_key_indices: vec![ENTRY_TYPE_INDEX, KEY_INDEX],
create_if_not_exists: true,
table_options: HashMap::new(),
};
@@ -127,14 +126,14 @@ impl SystemCatalogTable {
/// Create a stream of all entries inside system catalog table
pub async fn records(&self) -> Result<SendableRecordBatchStream> {
let full_projection = None;
let ctx = SessionContext::new();
let scan = self
.table
.scan(&full_projection, &[], None)
.await
.context(error::SystemCatalogTableScanSnafu)?;
let stream = scan
.execute(0, Arc::new(RuntimeEnv::default()))
.await
.execute(0, ctx.task_ctx())
.context(error::SystemCatalogTableScanExecSnafu)?;
Ok(stream)
}
@@ -162,7 +161,7 @@ fn build_system_catalog_schema() -> Schema {
),
ColumnSchema::new(
"timestamp".to_string(),
ConcreteDataType::timestamp_millis_datatype(),
ConcreteDataType::timestamp_millisecond_datatype(),
false,
)
.with_time_index(true),
@@ -173,12 +172,12 @@ fn build_system_catalog_schema() -> Schema {
),
ColumnSchema::new(
"gmt_created".to_string(),
ConcreteDataType::timestamp_millis_datatype(),
ConcreteDataType::timestamp_millisecond_datatype(),
false,
),
ColumnSchema::new(
"gmt_modified".to_string(),
ConcreteDataType::timestamp_millis_datatype(),
ConcreteDataType::timestamp_millisecond_datatype(),
false,
),
];
@@ -223,7 +222,7 @@ pub fn build_insert_request(entry_type: EntryType, key: &[u8], value: &[u8]) ->
// Timestamp in key part is intentionally left to 0
columns_values.insert(
"timestamp".to_string(),
Arc::new(TimestampVector::from_slice(&[Timestamp::from_millis(0)])) as _,
Arc::new(TimestampMillisecondVector::from_slice(&[0])) as _,
);
columns_values.insert(
@@ -231,18 +230,15 @@ pub fn build_insert_request(entry_type: EntryType, key: &[u8], value: &[u8]) ->
Arc::new(BinaryVector::from_slice(&[value])) as _,
);
let now = util::current_time_millis();
columns_values.insert(
"gmt_created".to_string(),
Arc::new(TimestampVector::from_slice(&[Timestamp::from_millis(
util::current_time_millis(),
)])) as _,
Arc::new(TimestampMillisecondVector::from_slice(&[now])) as _,
);
columns_values.insert(
"gmt_modified".to_string(),
Arc::new(TimestampVector::from_slice(&[Timestamp::from_millis(
util::current_time_millis(),
)])) as _,
Arc::new(TimestampMillisecondVector::from_slice(&[now])) as _,
);
InsertRequest {
@@ -384,7 +380,7 @@ mod tests {
use super::*;
#[test]
pub fn test_decode_catalog_enrty() {
pub fn test_decode_catalog_entry() {
let entry = decode_system_catalog(
Some(EntryType::Catalog as u8),
Some("some_catalog".as_bytes()),
@@ -456,7 +452,7 @@ mod tests {
pub async fn prepare_table_engine() -> (TempDir, TableEngineRef) {
let dir = TempDir::new("system-table-test").unwrap();
let store_dir = dir.path().to_string_lossy();
let accessor = opendal::services::fs::Builder::default()
let accessor = object_store::backend::fs::Builder::default()
.root(&store_dir)
.build()
.unwrap();

View File

@@ -26,9 +26,9 @@ use common_query::logical_plan::Expr;
use common_query::physical_plan::PhysicalPlanRef;
use common_recordbatch::error::Result as RecordBatchResult;
use common_recordbatch::{RecordBatch, RecordBatchStream};
use datatypes::prelude::{ConcreteDataType, VectorBuilder};
use datatypes::prelude::{ConcreteDataType, DataType};
use datatypes::schema::{ColumnSchema, Schema, SchemaRef};
use datatypes::value::Value;
use datatypes::value::ValueRef;
use datatypes::vectors::VectorRef;
use futures::Stream;
use snafu::ResultExt;
@@ -149,26 +149,33 @@ fn tables_to_record_batch(
engine: &str,
) -> Vec<VectorRef> {
let mut catalog_vec =
VectorBuilder::with_capacity(ConcreteDataType::string_datatype(), table_names.len());
ConcreteDataType::string_datatype().create_mutable_vector(table_names.len());
let mut schema_vec =
VectorBuilder::with_capacity(ConcreteDataType::string_datatype(), table_names.len());
ConcreteDataType::string_datatype().create_mutable_vector(table_names.len());
let mut table_name_vec =
VectorBuilder::with_capacity(ConcreteDataType::string_datatype(), table_names.len());
ConcreteDataType::string_datatype().create_mutable_vector(table_names.len());
let mut engine_vec =
VectorBuilder::with_capacity(ConcreteDataType::string_datatype(), table_names.len());
ConcreteDataType::string_datatype().create_mutable_vector(table_names.len());
for table_name in table_names {
catalog_vec.push(&Value::String(catalog_name.into()));
schema_vec.push(&Value::String(schema_name.into()));
table_name_vec.push(&Value::String(table_name.into()));
engine_vec.push(&Value::String(engine.into()));
// Safety: All these vectors are string type.
catalog_vec
.push_value_ref(ValueRef::String(catalog_name))
.unwrap();
schema_vec
.push_value_ref(ValueRef::String(schema_name))
.unwrap();
table_name_vec
.push_value_ref(ValueRef::String(&table_name))
.unwrap();
engine_vec.push_value_ref(ValueRef::String(engine)).unwrap();
}
vec![
catalog_vec.finish(),
schema_vec.finish(),
table_name_vec.finish(),
engine_vec.finish(),
catalog_vec.to_vector(),
schema_vec.to_vector(),
table_name_vec.to_vector(),
engine_vec.to_vector(),
]
}
@@ -340,9 +347,7 @@ fn build_schema_for_tables() -> Schema {
#[cfg(test)]
mod tests {
use common_catalog::consts::{DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME};
use common_query::physical_plan::RuntimeEnv;
use datatypes::arrow::array::Utf8Array;
use datatypes::arrow::datatypes::DataType;
use common_query::physical_plan::SessionContext;
use futures_util::StreamExt;
use table::table::numbers::NumbersTable;
@@ -366,57 +371,47 @@ mod tests {
let tables = Tables::new(catalog_list, "test_engine".to_string());
let tables_stream = tables.scan(&None, &[], None).await.unwrap();
let mut tables_stream = tables_stream
.execute(0, Arc::new(RuntimeEnv::default()))
.await
.unwrap();
let session_ctx = SessionContext::new();
let mut tables_stream = tables_stream.execute(0, session_ctx.task_ctx()).unwrap();
if let Some(t) = tables_stream.next().await {
let batch = t.unwrap().df_recordbatch;
let batch = t.unwrap();
assert_eq!(1, batch.num_rows());
assert_eq!(4, batch.num_columns());
assert_eq!(&DataType::Utf8, batch.column(0).data_type());
assert_eq!(&DataType::Utf8, batch.column(1).data_type());
assert_eq!(&DataType::Utf8, batch.column(2).data_type());
assert_eq!(&DataType::Utf8, batch.column(3).data_type());
assert_eq!(
ConcreteDataType::string_datatype(),
batch.column(0).data_type()
);
assert_eq!(
ConcreteDataType::string_datatype(),
batch.column(1).data_type()
);
assert_eq!(
ConcreteDataType::string_datatype(),
batch.column(2).data_type()
);
assert_eq!(
ConcreteDataType::string_datatype(),
batch.column(3).data_type()
);
assert_eq!(
"greptime",
batch
.column(0)
.as_any()
.downcast_ref::<Utf8Array<i32>>()
.unwrap()
.value(0)
batch.column(0).get_ref(0).as_string().unwrap().unwrap()
);
assert_eq!(
"public",
batch
.column(1)
.as_any()
.downcast_ref::<Utf8Array<i32>>()
.unwrap()
.value(0)
batch.column(1).get_ref(0).as_string().unwrap().unwrap()
);
assert_eq!(
"test_table",
batch
.column(2)
.as_any()
.downcast_ref::<Utf8Array<i32>>()
.unwrap()
.value(0)
batch.column(2).get_ref(0).as_string().unwrap().unwrap()
);
assert_eq!(
"test_engine",
batch
.column(3)
.as_any()
.downcast_ref::<Utf8Array<i32>>()
.unwrap()
.value(0)
batch.column(3).get_ref(0).as_string().unwrap().unwrap()
);
} else {
panic!("Record batch should not be empty!")

View File

@@ -0,0 +1,132 @@
// Copyright 2022 Greptime Team
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#[cfg(test)]
mod tests {
use std::sync::Arc;
use catalog::local::LocalCatalogManager;
use catalog::{CatalogManager, RegisterTableRequest};
use common_catalog::consts::{DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME};
use common_telemetry::{error, info};
use mito::config::EngineConfig;
use table::table::numbers::NumbersTable;
use table::TableRef;
use tokio::sync::Mutex;
async fn create_local_catalog_manager() -> Result<LocalCatalogManager, catalog::error::Error> {
let (_dir, object_store) =
mito::table::test_util::new_test_object_store("setup_mock_engine_and_table").await;
let mock_engine = Arc::new(mito::table::test_util::MockMitoEngine::new(
EngineConfig::default(),
mito::table::test_util::MockEngine::default(),
object_store,
));
let catalog_manager = LocalCatalogManager::try_new(mock_engine).await.unwrap();
catalog_manager.start().await?;
Ok(catalog_manager)
}
#[tokio::test]
async fn test_duplicate_register() {
let catalog_manager = create_local_catalog_manager().await.unwrap();
let request = RegisterTableRequest {
catalog: DEFAULT_CATALOG_NAME.to_string(),
schema: DEFAULT_SCHEMA_NAME.to_string(),
table_name: "test_table".to_string(),
table_id: 42,
table: Arc::new(NumbersTable::new(42)),
};
assert!(catalog_manager
.register_table(request.clone())
.await
.unwrap());
// register table with same table id will succeed with 0 as return val.
assert!(!catalog_manager.register_table(request).await.unwrap());
let err = catalog_manager
.register_table(RegisterTableRequest {
catalog: DEFAULT_CATALOG_NAME.to_string(),
schema: DEFAULT_SCHEMA_NAME.to_string(),
table_name: "test_table".to_string(),
table_id: 43,
table: Arc::new(NumbersTable::new(43)),
})
.await
.unwrap_err();
assert!(
err.to_string()
.contains("Table `greptime.public.test_table` already exists"),
"Actual error message: {}",
err
);
}
#[test]
fn test_concurrent_register() {
common_telemetry::init_default_ut_logging();
let rt = Arc::new(tokio::runtime::Builder::new_multi_thread().build().unwrap());
let catalog_manager =
Arc::new(rt.block_on(async { create_local_catalog_manager().await.unwrap() }));
let succeed: Arc<Mutex<Option<TableRef>>> = Arc::new(Mutex::new(None));
let mut handles = Vec::with_capacity(8);
for i in 0..8 {
let catalog = catalog_manager.clone();
let succeed = succeed.clone();
let handle = rt.spawn(async move {
let table_id = 42 + i;
let table = Arc::new(NumbersTable::new(table_id));
let req = RegisterTableRequest {
catalog: DEFAULT_CATALOG_NAME.to_string(),
schema: DEFAULT_SCHEMA_NAME.to_string(),
table_name: "test_table".to_string(),
table_id,
table: table.clone(),
};
match catalog.register_table(req).await {
Ok(res) => {
if res {
let mut succeed = succeed.lock().await;
info!("Successfully registered table: {}", table_id);
*succeed = Some(table);
}
}
Err(_) => {
error!("Failed to register table {}", table_id);
}
}
});
handles.push(handle);
}
rt.block_on(async move {
for handle in handles {
handle.await.unwrap();
}
let guard = succeed.lock().await;
let table = guard.as_ref().unwrap();
let table_registered = catalog_manager
.table(DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME, "test_table")
.unwrap()
.unwrap();
assert_eq!(
table_registered.table_info().ident.table_id,
table.table_info().ident.table_id
);
});
}
}

View File

@@ -217,7 +217,7 @@ impl TableEngine for MockTableEngine {
&self,
_ctx: &EngineContext,
_request: DropTableRequest,
) -> table::Result<()> {
) -> table::Result<bool> {
unimplemented!()
}
}

View File

@@ -22,12 +22,12 @@ mod tests {
use std::collections::HashSet;
use std::sync::Arc;
use catalog::helper::{CatalogKey, CatalogValue, SchemaKey, SchemaValue};
use catalog::remote::{
KvBackend, KvBackendRef, RemoteCatalogManager, RemoteCatalogProvider, RemoteSchemaProvider,
};
use catalog::{CatalogList, CatalogManager, RegisterTableRequest};
use common_catalog::consts::{DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME};
use common_catalog::{CatalogKey, CatalogValue, SchemaKey, SchemaValue};
use datatypes::schema::Schema;
use futures_util::StreamExt;
use table::engine::{EngineContext, TableEngineRef};
@@ -202,7 +202,7 @@ mod tests {
table_id,
table,
};
assert_eq!(1, catalog_manager.register_table(reg_req).await.unwrap());
assert!(catalog_manager.register_table(reg_req).await.unwrap());
assert_eq!(
HashSet::from([table_name, "numbers".to_string()]),
default_schema
@@ -287,7 +287,7 @@ mod tests {
.register_schema(schema_name.clone(), schema.clone())
.expect("Register schema should not fail");
assert!(prev.is_none());
assert_eq!(1, catalog_manager.register_table(reg_req).await.unwrap());
assert!(catalog_manager.register_table(reg_req).await.unwrap());
assert_eq!(
HashSet::from([schema_name.clone()]),

View File

@@ -11,13 +11,11 @@ async-stream = "0.3"
common-base = { path = "../common/base" }
common-error = { path = "../common/error" }
common-grpc = { path = "../common/grpc" }
common-grpc-expr = { path = "../common/grpc-expr" }
common-query = { path = "../common/query" }
common-recordbatch = { path = "../common/recordbatch" }
common-insert = { path = "../common/insert" }
common-time = { path = "../common/time" }
datafusion = { git = "https://github.com/apache/arrow-datafusion.git", branch = "arrow2", features = [
"simd",
] }
datafusion = "14.0.0"
datatypes = { path = "../datatypes" }
enum_dispatch = "0.3"
parking_lot = "0.12"

View File

@@ -12,11 +12,9 @@
// See the License for the specific language governing permissions and
// limitations under the License.
use std::collections::HashMap;
use api::v1::codec::InsertBatch;
use api::v1::*;
use client::{Client, Database};
fn main() {
tracing::subscriber::set_global_default(tracing_subscriber::FmtSubscriber::builder().finish())
.unwrap();
@@ -29,21 +27,21 @@ async fn run() {
let client = Client::with_urls(vec!["127.0.0.1:3001"]);
let db = Database::new("greptime", client);
let (columns, row_count) = insert_data();
let expr = InsertExpr {
schema_name: "public".to_string(),
table_name: "demo".to_string(),
expr: Some(insert_expr::Expr::Values(insert_expr::Values {
values: insert_batches(),
})),
options: HashMap::default(),
region_number: 0,
columns,
row_count,
};
db.insert(expr).await.unwrap();
}
fn insert_batches() -> Vec<Vec<u8>> {
fn insert_data() -> (Vec<Column>, u32) {
const SEMANTIC_TAG: i32 = 0;
const SEMANTIC_FEILD: i32 = 1;
const SEMANTIC_FIELD: i32 = 1;
const SEMANTIC_TS: i32 = 2;
let row_count = 4;
@@ -71,7 +69,7 @@ fn insert_batches() -> Vec<Vec<u8>> {
};
let cpu_column = Column {
column_name: "cpu".to_string(),
semantic_type: SEMANTIC_FEILD,
semantic_type: SEMANTIC_FIELD,
values: Some(cpu_vals),
null_mask: vec![2],
..Default::default()
@@ -83,7 +81,7 @@ fn insert_batches() -> Vec<Vec<u8>> {
};
let mem_column = Column {
column_name: "memory".to_string(),
semantic_type: SEMANTIC_FEILD,
semantic_type: SEMANTIC_FIELD,
values: Some(mem_vals),
null_mask: vec![4],
..Default::default()
@@ -101,9 +99,8 @@ fn insert_batches() -> Vec<Vec<u8>> {
..Default::default()
};
let insert_batch = InsertBatch {
columns: vec![host_column, cpu_column, mem_column, ts_column],
(
vec![host_column, cpu_column, mem_column, ts_column],
row_count,
};
vec![insert_batch.into()]
)
}

View File

@@ -41,7 +41,7 @@ async fn run() {
column_defs: vec![
ColumnDef {
name: "timestamp".to_string(),
datatype: ColumnDataType::Timestamp as i32,
datatype: ColumnDataType::TimestampMillisecond as i32,
is_nullable: false,
default_constraint: None,
},

View File

@@ -1,51 +0,0 @@
// Copyright 2022 Greptime Team
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
use std::sync::Arc;
use client::{Client, Database};
use common_grpc::MockExecution;
use datafusion::physical_plan::expressions::Column;
use datafusion::physical_plan::projection::ProjectionExec;
use datafusion::physical_plan::{ExecutionPlan, PhysicalExpr};
use tracing::{event, Level};
fn main() {
tracing::subscriber::set_global_default(tracing_subscriber::FmtSubscriber::builder().finish())
.unwrap();
run();
}
#[tokio::main]
async fn run() {
let client = Client::with_urls(vec!["127.0.0.1:3001"]);
let db = Database::new("greptime", client);
let physical = mock_physical_plan();
let result = db.physical_plan(physical, None).await;
event!(Level::INFO, "result: {:#?}", result);
}
fn mock_physical_plan() -> Arc<dyn ExecutionPlan> {
let id_expr = Arc::new(Column::new("id", 0)) as Arc<dyn PhysicalExpr>;
let age_expr = Arc::new(Column::new("age", 2)) as Arc<dyn PhysicalExpr>;
let expr = vec![(id_expr, "id".to_string()), (age_expr, "age".to_string())];
let input =
Arc::new(MockExecution::new("mock_input_exec".to_string())) as Arc<dyn ExecutionPlan>;
let projection = ProjectionExec::try_new(expr, input).unwrap();
Arc::new(projection)
}

View File

@@ -58,7 +58,19 @@ impl Admin {
header: Some(header),
expr: Some(admin_expr::Expr::Alter(expr)),
};
Ok(self.do_requests(vec![expr]).await?.remove(0))
self.do_request(expr).await
}
pub async fn drop_table(&self, expr: DropTableExpr) -> Result<AdminResult> {
let header = ExprHeader {
version: PROTOCOL_VERSION,
};
let expr = AdminExpr {
header: Some(header),
expr: Some(admin_expr::Expr::DropTable(expr)),
};
self.do_request(expr).await
}
/// Invariants: the lengths of input vec (`Vec<AdminExpr>`) and output vec (`Vec<AdminResult>`) are equal.

View File

@@ -18,22 +18,17 @@ use api::v1::codec::SelectResult as GrpcSelectResult;
use api::v1::column::SemanticType;
use api::v1::{
object_expr, object_result, select_expr, DatabaseRequest, ExprHeader, InsertExpr,
MutateResult as GrpcMutateResult, ObjectExpr, ObjectResult as GrpcObjectResult, PhysicalPlan,
SelectExpr,
MutateResult as GrpcMutateResult, ObjectExpr, ObjectResult as GrpcObjectResult, SelectExpr,
};
use common_error::status_code::StatusCode;
use common_grpc::{AsExcutionPlan, DefaultAsPlanImpl};
use common_insert::column_to_vector;
use common_grpc_expr::column_to_vector;
use common_query::Output;
use common_recordbatch::{RecordBatch, RecordBatches};
use datafusion::physical_plan::ExecutionPlan;
use datatypes::prelude::*;
use datatypes::schema::{ColumnSchema, Schema};
use snafu::{ensure, OptionExt, ResultExt};
use crate::error::{
ColumnToVectorSnafu, ConvertSchemaSnafu, DatanodeSnafu, DecodeSelectSnafu, EncodePhysicalSnafu,
};
use crate::error::{ColumnToVectorSnafu, ConvertSchemaSnafu, DatanodeSnafu, DecodeSelectSnafu};
use crate::{error, Client, Result};
pub const PROTOCOL_VERSION: u32 = 1;
@@ -94,24 +89,6 @@ impl Database {
self.do_select(select_expr).await
}
pub async fn physical_plan(
&self,
physical: Arc<dyn ExecutionPlan>,
original_ql: Option<String>,
) -> Result<ObjectResult> {
let plan = DefaultAsPlanImpl::try_from_physical_plan(physical.clone())
.context(EncodePhysicalSnafu { physical })?
.bytes;
let original_ql = original_ql.unwrap_or_default();
let select_expr = SelectExpr {
expr: Some(select_expr::Expr::PhysicalPlan(PhysicalPlan {
original_ql: original_ql.into_bytes(),
plan,
})),
};
self.do_select(select_expr).await
}
pub async fn logical_plan(&self, logical_plan: Vec<u8>) -> Result<ObjectResult> {
let select_expr = SelectExpr {
expr: Some(select_expr::Expr::LogicalPlan(logical_plan)),
@@ -341,12 +318,11 @@ mod tests {
fn create_test_column(vector: VectorRef) -> Column {
let wrapper: ColumnDataTypeWrapper = vector.data_type().try_into().unwrap();
let array = vector.to_arrow_array();
Column {
column_name: "test".to_string(),
semantic_type: 1,
values: Some(values(&[array.clone()]).unwrap()),
null_mask: null_mask(&vec![array], vector.len()),
values: Some(values(&[vector.clone()]).unwrap()),
null_mask: null_mask(&[vector.clone()], vector.len()),
datatype: wrapper.datatype() as i32,
}
}

View File

@@ -103,7 +103,7 @@ pub enum Error {
#[snafu(display("Failed to convert column to vector, source: {}", source))]
ColumnToVector {
#[snafu(backtrace)]
source: common_insert::error::Error,
source: common_grpc_expr::error::Error,
},
}

View File

@@ -10,6 +10,7 @@ name = "greptime"
path = "src/bin/greptime.rs"
[dependencies]
anymap = "1.0.0-beta.2"
clap = { version = "3.1", features = ["derive"] }
common-error = { path = "../common/error" }
common-telemetry = { path = "../common/telemetry", features = [
@@ -18,8 +19,10 @@ common-telemetry = { path = "../common/telemetry", features = [
datanode = { path = "../datanode" }
frontend = { path = "../frontend" }
futures = "0.3"
meta-client = { path = "../meta-client" }
meta-srv = { path = "../meta-srv" }
serde = "1.0"
servers = { path = "../servers" }
snafu = { version = "0.7", features = ["backtraces"] }
tokio = { version = "1.18", features = ["full"] }
toml = "0.5"
@@ -27,3 +30,6 @@ toml = "0.5"
[dev-dependencies]
serde = "1.0"
tempdir = "0.3"
[build-dependencies]
build-data = "0.1.3"

29
src/cmd/build.rs Normal file
View File

@@ -0,0 +1,29 @@
// Copyright 2022 Greptime Team
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
const DEFAULT_VALUE: &str = "unknown";
fn main() {
println!(
"cargo:rustc-env=GIT_COMMIT={}",
build_data::get_git_commit().unwrap_or_else(|_| DEFAULT_VALUE.to_string())
);
println!(
"cargo:rustc-env=GIT_BRANCH={}",
build_data::get_git_branch().unwrap_or_else(|_| DEFAULT_VALUE.to_string())
);
println!(
"cargo:rustc-env=GIT_DIRTY={}",
build_data::get_git_dirty().map_or(DEFAULT_VALUE.to_string(), |v| v.to_string())
);
}

View File

@@ -20,7 +20,7 @@ use cmd::{datanode, frontend, metasrv, standalone};
use common_telemetry::logging::{error, info};
#[derive(Parser)]
#[clap(name = "greptimedb")]
#[clap(name = "greptimedb", version = print_version())]
struct Command {
#[clap(long, default_value = "/tmp/greptimedb/logs")]
log_dir: String,
@@ -70,6 +70,19 @@ impl fmt::Display for SubCommand {
}
}
fn print_version() -> &'static str {
concat!(
"\nbranch: ",
env!("GIT_BRANCH"),
"\ncommit: ",
env!("GIT_COMMIT"),
"\ndirty: ",
env!("GIT_DIRTY"),
"\nversion: ",
env!("CARGO_PKG_VERSION")
)
}
#[tokio::main]
async fn main() -> Result<()> {
let cmd = Command::parse();

View File

@@ -14,8 +14,9 @@
use clap::Parser;
use common_telemetry::logging;
use datanode::datanode::{Datanode, DatanodeOptions};
use frontend::frontend::Mode;
use datanode::datanode::{Datanode, DatanodeOptions, ObjectStoreConfig};
use meta_client::MetaClientOpts;
use servers::Mode;
use snafu::ResultExt;
use crate::error::{Error, MissingConfigSnafu, Result, StartDatanodeSnafu};
@@ -46,7 +47,7 @@ impl SubCommand {
}
}
#[derive(Debug, Parser)]
#[derive(Debug, Parser, Default)]
struct StartCommand {
#[clap(long)]
node_id: Option<u64>,
@@ -58,6 +59,10 @@ struct StartCommand {
metasrv_addr: Option<String>,
#[clap(short, long)]
config_file: Option<String>,
#[clap(long)]
data_dir: Option<String>,
#[clap(long)]
wal_dir: Option<String>,
}
impl StartCommand {
@@ -98,7 +103,13 @@ impl TryFrom<StartCommand> for DatanodeOptions {
}
if let Some(meta_addr) = cmd.metasrv_addr {
opts.meta_client_opts.metasrv_addr = meta_addr;
opts.meta_client_opts
.get_or_insert_with(MetaClientOpts::default)
.metasrv_addrs = meta_addr
.split(',')
.map(&str::trim)
.map(&str::to_string)
.collect::<_>();
opts.mode = Mode::Distributed;
}
@@ -108,6 +119,14 @@ impl TryFrom<StartCommand> for DatanodeOptions {
}
.fail();
}
if let Some(data_dir) = cmd.data_dir {
opts.storage = ObjectStoreConfig::File { data_dir };
}
if let Some(wal_dir) = cmd.wal_dir {
opts.wal_dir = wal_dir;
}
Ok(opts)
}
}
@@ -117,39 +136,41 @@ mod tests {
use std::assert_matches::assert_matches;
use datanode::datanode::ObjectStoreConfig;
use frontend::frontend::Mode;
use servers::Mode;
use super::*;
#[test]
fn test_read_from_config_file() {
let cmd = StartCommand {
node_id: None,
rpc_addr: None,
mysql_addr: None,
metasrv_addr: None,
config_file: Some(format!(
"{}/../../config/datanode.example.toml",
std::env::current_dir().unwrap().as_path().to_str().unwrap()
)),
..Default::default()
};
let options: DatanodeOptions = cmd.try_into().unwrap();
assert_eq!("127.0.0.1:3001".to_string(), options.rpc_addr);
assert_eq!("/tmp/greptimedb/wal".to_string(), options.wal_dir);
assert_eq!("127.0.0.1:3306".to_string(), options.mysql_addr);
assert_eq!("127.0.0.1:4406".to_string(), options.mysql_addr);
assert_eq!(4, options.mysql_runtime_size);
assert_eq!(
"1.1.1.1:3002".to_string(),
options.meta_client_opts.metasrv_addr
);
assert_eq!(5000, options.meta_client_opts.connect_timeout_millis);
assert_eq!(3000, options.meta_client_opts.timeout_millis);
assert!(!options.meta_client_opts.tcp_nodelay);
let MetaClientOpts {
metasrv_addrs: metasrv_addr,
timeout_millis,
connect_timeout_millis,
tcp_nodelay,
} = options.meta_client_opts.unwrap();
assert_eq!(vec!["127.0.0.1:3002".to_string()], metasrv_addr);
assert_eq!(5000, connect_timeout_millis);
assert_eq!(3000, timeout_millis);
assert!(!tcp_nodelay);
match options.storage {
ObjectStoreConfig::File { data_dir } => {
assert_eq!("/tmp/greptimedb/data/".to_string(), data_dir)
}
ObjectStoreConfig::S3 { .. } => unreachable!(),
};
}
@@ -157,44 +178,30 @@ mod tests {
fn test_try_from_cmd() {
assert_eq!(
Mode::Standalone,
DatanodeOptions::try_from(StartCommand {
node_id: None,
rpc_addr: None,
mysql_addr: None,
metasrv_addr: None,
config_file: None
})
.unwrap()
.mode
DatanodeOptions::try_from(StartCommand::default())
.unwrap()
.mode
);
let mode = DatanodeOptions::try_from(StartCommand {
node_id: Some(42),
rpc_addr: None,
mysql_addr: None,
metasrv_addr: Some("127.0.0.1:3002".to_string()),
config_file: None,
..Default::default()
})
.unwrap()
.mode;
assert_matches!(mode, Mode::Distributed);
assert!(DatanodeOptions::try_from(StartCommand {
node_id: None,
rpc_addr: None,
mysql_addr: None,
metasrv_addr: Some("127.0.0.1:3002".to_string()),
config_file: None,
..Default::default()
})
.is_err());
// Providing node_id but leave metasrv_addr absent is ok since metasrv_addr has default value
DatanodeOptions::try_from(StartCommand {
node_id: Some(42),
rpc_addr: None,
mysql_addr: None,
metasrv_addr: None,
config_file: None,
..Default::default()
})
.unwrap();
}
@@ -202,17 +209,23 @@ mod tests {
#[test]
fn test_merge_config() {
let dn_opts = DatanodeOptions::try_from(StartCommand {
node_id: None,
rpc_addr: None,
mysql_addr: None,
metasrv_addr: None,
config_file: Some(format!(
"{}/../../config/datanode.example.toml",
std::env::current_dir().unwrap().as_path().to_str().unwrap()
)),
..Default::default()
})
.unwrap();
assert_eq!(Some(42), dn_opts.node_id);
assert_eq!("1.1.1.1:3002", dn_opts.meta_client_opts.metasrv_addr);
let MetaClientOpts {
metasrv_addrs: metasrv_addr,
timeout_millis,
connect_timeout_millis,
tcp_nodelay,
} = dn_opts.meta_client_opts.unwrap();
assert_eq!(vec!["127.0.0.1:3002".to_string()], metasrv_addr);
assert_eq!(3000, timeout_millis);
assert_eq!(5000, connect_timeout_millis);
assert!(!tcp_nodelay);
}
}

View File

@@ -25,12 +25,6 @@ pub enum Error {
source: datanode::error::Error,
},
#[snafu(display("Failed to build frontend, source: {}", source))]
BuildFrontend {
#[snafu(backtrace)]
source: frontend::error::Error,
},
#[snafu(display("Failed to start frontend, source: {}", source))]
StartFrontend {
#[snafu(backtrace)]
@@ -61,6 +55,12 @@ pub enum Error {
#[snafu(display("Illegal config: {}", msg))]
IllegalConfig { msg: String, backtrace: Backtrace },
#[snafu(display("Illegal auth config: {}", source))]
IllegalAuthConfig {
#[snafu(backtrace)]
source: servers::auth::Error,
},
}
pub type Result<T> = std::result::Result<T, Error>;
@@ -75,7 +75,7 @@ impl ErrorExt for Error {
StatusCode::InvalidArguments
}
Error::IllegalConfig { .. } => StatusCode::InvalidArguments,
Error::BuildFrontend { source, .. } => source.status_code(),
Error::IllegalAuthConfig { .. } => StatusCode::InvalidArguments,
}
}
@@ -97,10 +97,7 @@ mod tests {
#[test]
fn test_start_node_error() {
fn throw_datanode_error() -> StdResult<datanode::error::Error> {
datanode::error::MissingFieldSnafu {
field: "test_field",
}
.fail()
datanode::error::MissingNodeIdSnafu {}.fail()
}
let e = throw_datanode_error()

View File

@@ -12,17 +12,23 @@
// See the License for the specific language governing permissions and
// limitations under the License.
use anymap::AnyMap;
use clap::Parser;
use frontend::frontend::{Frontend, FrontendOptions, Mode};
use frontend::frontend::{Frontend, FrontendOptions};
use frontend::grpc::GrpcOptions;
use frontend::influxdb::InfluxdbOptions;
use frontend::instance::Instance;
use frontend::mysql::MysqlOptions;
use frontend::opentsdb::OpentsdbOptions;
use frontend::postgres::PostgresOptions;
use meta_client::MetaClientOpts;
use servers::auth::UserProviderRef;
use servers::http::HttpOptions;
use servers::tls::{TlsMode, TlsOption};
use servers::{auth, Mode};
use snafu::ResultExt;
use crate::error::{self, Result};
use crate::error::{self, IllegalAuthConfigSnafu, Result};
use crate::toml_loader;
#[derive(Parser)]
@@ -68,21 +74,41 @@ pub struct StartCommand {
influxdb_enable: Option<bool>,
#[clap(long)]
metasrv_addr: Option<String>,
#[clap(long)]
tls_mode: Option<TlsMode>,
#[clap(long)]
tls_cert_path: Option<String>,
#[clap(long)]
tls_key_path: Option<String>,
#[clap(long)]
user_provider: Option<String>,
}
impl StartCommand {
async fn run(self) -> Result<()> {
let plugins = load_frontend_plugins(&self.user_provider)?;
let opts: FrontendOptions = self.try_into()?;
let mut frontend = Frontend::new(
opts.clone(),
Instance::try_new(&opts)
Instance::try_new_distributed(&opts)
.await
.context(error::StartFrontendSnafu)?,
plugins,
);
frontend.start().await.context(error::StartFrontendSnafu)
}
}
pub fn load_frontend_plugins(user_provider: &Option<String>) -> Result<AnyMap> {
let mut plugins = AnyMap::new();
if let Some(provider) = user_provider {
let provider = auth::user_provider_from_option(provider).context(IllegalAuthConfigSnafu)?;
plugins.insert::<UserProviderRef>(provider);
}
Ok(plugins)
}
impl TryFrom<StartCommand> for FrontendOptions {
type Error = error::Error;
@@ -93,8 +119,13 @@ impl TryFrom<StartCommand> for FrontendOptions {
FrontendOptions::default()
};
let tls_option = TlsOption::new(cmd.tls_mode, cmd.tls_cert_path, cmd.tls_key_path);
if let Some(addr) = cmd.http_addr {
opts.http_addr = Some(addr);
opts.http_options = Some(HttpOptions {
addr,
..Default::default()
});
}
if let Some(addr) = cmd.grpc_addr {
opts.grpc_options = Some(GrpcOptions {
@@ -105,12 +136,14 @@ impl TryFrom<StartCommand> for FrontendOptions {
if let Some(addr) = cmd.mysql_addr {
opts.mysql_options = Some(MysqlOptions {
addr,
tls: tls_option.clone(),
..Default::default()
});
}
if let Some(addr) = cmd.postgres_addr {
opts.postgres_options = Some(PostgresOptions {
addr,
tls: tls_option,
..Default::default()
});
}
@@ -124,13 +157,13 @@ impl TryFrom<StartCommand> for FrontendOptions {
opts.influxdb_options = Some(InfluxdbOptions { enable });
}
if let Some(metasrv_addr) = cmd.metasrv_addr {
opts.metasrv_addr = Some(
metasrv_addr
.split(',')
.into_iter()
.map(|x| x.trim().to_string())
.collect::<Vec<String>>(),
);
opts.meta_client_opts
.get_or_insert_with(MetaClientOpts::default)
.metasrv_addrs = metasrv_addr
.split(',')
.map(&str::trim)
.map(&str::to_string)
.collect::<Vec<_>>();
opts.mode = Mode::Distributed;
}
Ok(opts)
@@ -139,6 +172,10 @@ impl TryFrom<StartCommand> for FrontendOptions {
#[cfg(test)]
mod tests {
use std::time::Duration;
use servers::auth::{Identity, Password, UserProviderRef};
use super::*;
#[test]
@@ -152,10 +189,14 @@ mod tests {
influxdb_enable: Some(false),
config_file: None,
metasrv_addr: None,
tls_mode: None,
tls_cert_path: None,
tls_key_path: None,
user_provider: None,
};
let opts: FrontendOptions = command.try_into().unwrap();
assert_eq!(opts.http_addr, Some("127.0.0.1:1234".to_string()));
assert_eq!(opts.http_options.as_ref().unwrap().addr, "127.0.0.1:1234");
assert_eq!(opts.mysql_options.as_ref().unwrap().addr, "127.0.0.1:5678");
assert_eq!(
opts.postgres_options.as_ref().unwrap().addr,
@@ -186,4 +227,66 @@ mod tests {
assert!(!opts.influxdb_options.unwrap().enable);
}
#[test]
fn test_read_from_config_file() {
let command = StartCommand {
http_addr: None,
grpc_addr: None,
mysql_addr: None,
postgres_addr: None,
opentsdb_addr: None,
influxdb_enable: None,
config_file: Some(format!(
"{}/../../config/frontend.example.toml",
std::env::current_dir().unwrap().as_path().to_str().unwrap()
)),
metasrv_addr: None,
tls_mode: None,
tls_cert_path: None,
tls_key_path: None,
user_provider: None,
};
let fe_opts = FrontendOptions::try_from(command).unwrap();
assert_eq!(Mode::Distributed, fe_opts.mode);
assert_eq!(
"127.0.0.1:4000".to_string(),
fe_opts.http_options.as_ref().unwrap().addr
);
assert_eq!(
Duration::from_secs(30),
fe_opts.http_options.as_ref().unwrap().timeout
);
}
#[tokio::test]
async fn test_try_from_start_command_to_anymap() {
let command = StartCommand {
http_addr: None,
grpc_addr: None,
mysql_addr: None,
postgres_addr: None,
opentsdb_addr: None,
influxdb_enable: None,
config_file: None,
metasrv_addr: None,
tls_mode: None,
tls_cert_path: None,
tls_key_path: None,
user_provider: Some("static_user_provider:cmd:test=test".to_string()),
};
let plugins = load_frontend_plugins(&command.user_provider);
assert!(plugins.is_ok());
let plugins = plugins.unwrap();
let provider = plugins.get::<UserProviderRef>();
assert!(provider.is_some());
let provider = provider.unwrap();
let result = provider
.auth(Identity::UserId("test", None), Password::PlainText("test"))
.await;
assert!(result.is_ok());
}
}

View File

@@ -12,11 +12,12 @@
// See the License for the specific language governing permissions and
// limitations under the License.
use anymap::AnyMap;
use clap::Parser;
use common_telemetry::info;
use datanode::datanode::{Datanode, DatanodeOptions, ObjectStoreConfig};
use datanode::instance::InstanceRef;
use frontend::frontend::{Frontend, FrontendOptions, Mode};
use frontend::frontend::{Frontend, FrontendOptions};
use frontend::grpc::GrpcOptions;
use frontend::influxdb::InfluxdbOptions;
use frontend::instance::Instance as FeInstance;
@@ -25,12 +26,13 @@ use frontend::opentsdb::OpentsdbOptions;
use frontend::postgres::PostgresOptions;
use frontend::prometheus::PrometheusOptions;
use serde::{Deserialize, Serialize};
use servers::http::HttpOptions;
use servers::tls::{TlsMode, TlsOption};
use servers::Mode;
use snafu::ResultExt;
use tokio::try_join;
use crate::error::{
BuildFrontendSnafu, Error, IllegalConfigSnafu, Result, StartDatanodeSnafu, StartFrontendSnafu,
};
use crate::error::{Error, IllegalConfigSnafu, Result, StartDatanodeSnafu, StartFrontendSnafu};
use crate::frontend::load_frontend_plugins;
use crate::toml_loader;
#[derive(Parser)]
@@ -60,7 +62,7 @@ impl SubCommand {
#[derive(Clone, Debug, Serialize, Deserialize)]
pub struct StandaloneOptions {
pub http_addr: Option<String>,
pub http_options: Option<HttpOptions>,
pub grpc_options: Option<GrpcOptions>,
pub mysql_options: Option<MysqlOptions>,
pub postgres_options: Option<PostgresOptions>,
@@ -70,14 +72,13 @@ pub struct StandaloneOptions {
pub mode: Mode,
pub wal_dir: String,
pub storage: ObjectStoreConfig,
pub datanode_mysql_addr: String,
pub datanode_mysql_runtime_size: usize,
pub enable_memory_catalog: bool,
}
impl Default for StandaloneOptions {
fn default() -> Self {
Self {
http_addr: Some("127.0.0.1:4000".to_string()),
http_options: Some(HttpOptions::default()),
grpc_options: Some(GrpcOptions::default()),
mysql_options: Some(MysqlOptions::default()),
postgres_options: Some(PostgresOptions::default()),
@@ -87,8 +88,7 @@ impl Default for StandaloneOptions {
mode: Mode::Standalone,
wal_dir: "/tmp/greptimedb/wal".to_string(),
storage: ObjectStoreConfig::default(),
datanode_mysql_addr: "127.0.0.1:3306".to_string(),
datanode_mysql_runtime_size: 4,
enable_memory_catalog: false,
}
}
}
@@ -96,7 +96,7 @@ impl Default for StandaloneOptions {
impl StandaloneOptions {
fn frontend_options(self) -> FrontendOptions {
FrontendOptions {
http_addr: self.http_addr,
http_options: self.http_options,
grpc_options: self.grpc_options,
mysql_options: self.mysql_options,
postgres_options: self.postgres_options,
@@ -104,8 +104,7 @@ impl StandaloneOptions {
influxdb_options: self.influxdb_options,
prometheus_options: self.prometheus_options,
mode: self.mode,
datanode_rpc_addr: "127.0.0.1:3001".to_string(),
metasrv_addr: None,
meta_client_opts: None,
}
}
@@ -113,8 +112,7 @@ impl StandaloneOptions {
DatanodeOptions {
wal_dir: self.wal_dir,
storage: self.storage,
mysql_addr: self.datanode_mysql_addr,
mysql_runtime_size: self.datanode_mysql_runtime_size,
enable_memory_catalog: self.enable_memory_catalog,
..Default::default()
}
}
@@ -136,18 +134,31 @@ struct StartCommand {
influxdb_enable: bool,
#[clap(short, long)]
config_file: Option<String>,
#[clap(short = 'm', long = "memory-catalog")]
enable_memory_catalog: bool,
#[clap(long)]
tls_mode: Option<TlsMode>,
#[clap(long)]
tls_cert_path: Option<String>,
#[clap(long)]
tls_key_path: Option<String>,
#[clap(long)]
user_provider: Option<String>,
}
impl StartCommand {
async fn run(self) -> Result<()> {
let enable_memory_catalog = self.enable_memory_catalog;
let config_file = self.config_file.clone();
let plugins = load_frontend_plugins(&self.user_provider)?;
let fe_opts = FrontendOptions::try_from(self)?;
let dn_opts: DatanodeOptions = {
let opts: StandaloneOptions = if let Some(path) = config_file {
let mut opts: StandaloneOptions = if let Some(path) = config_file {
toml_loader::from_file!(&path)?
} else {
StandaloneOptions::default()
};
opts.enable_memory_catalog = enable_memory_catalog;
opts.datanode_options()
};
@@ -159,13 +170,16 @@ impl StartCommand {
let mut datanode = Datanode::new(dn_opts.clone())
.await
.context(StartDatanodeSnafu)?;
let mut frontend = build_frontend(fe_opts, &dn_opts, datanode.get_instance()).await?;
let mut frontend = build_frontend(fe_opts, plugins, datanode.get_instance()).await?;
try_join!(
async { datanode.start().await.context(StartDatanodeSnafu) },
async { frontend.start().await.context(StartFrontendSnafu) }
)?;
// Start datanode instance before starting services, to avoid requests come in before internal components are started.
datanode
.start_instance()
.await
.context(StartDatanodeSnafu)?;
info!("Datanode instance started");
frontend.start().await.context(StartFrontendSnafu)?;
Ok(())
}
}
@@ -173,20 +187,12 @@ impl StartCommand {
/// Build frontend instance in standalone mode
async fn build_frontend(
fe_opts: FrontendOptions,
dn_opts: &DatanodeOptions,
plugins: AnyMap,
datanode_instance: InstanceRef,
) -> Result<Frontend<FeInstance>> {
let grpc_server_addr = &dn_opts.rpc_addr;
info!(
"Build frontend with datanode gRPC addr: {}",
grpc_server_addr
);
let mut frontend_instance = FeInstance::try_new(&fe_opts)
.await
.context(BuildFrontendSnafu)?;
frontend_instance.set_catalog_manager(datanode_instance.catalog_manager().clone());
let mut frontend_instance = FeInstance::new_standalone(datanode_instance.clone());
frontend_instance.set_script_handler(datanode_instance);
Ok(Frontend::new(fe_opts, frontend_instance))
Ok(Frontend::new(fe_opts, frontend_instance, plugins))
}
impl TryFrom<StartCommand> for FrontendOptions {
@@ -204,7 +210,10 @@ impl TryFrom<StartCommand> for FrontendOptions {
opts.mode = Mode::Standalone;
if let Some(addr) = cmd.http_addr {
opts.http_addr = Some(addr);
opts.http_options = Some(HttpOptions {
addr,
..Default::default()
});
}
if let Some(addr) = cmd.rpc_addr {
// frontend grpc addr conflict with datanode default grpc addr
@@ -248,12 +257,28 @@ impl TryFrom<StartCommand> for FrontendOptions {
opts.influxdb_options = Some(InfluxdbOptions { enable: true });
}
let tls_option = TlsOption::new(cmd.tls_mode, cmd.tls_cert_path, cmd.tls_key_path);
if let Some(mut mysql_options) = opts.mysql_options {
mysql_options.tls = tls_option.clone();
opts.mysql_options = Some(mysql_options);
}
if let Some(mut postgres_options) = opts.postgres_options {
postgres_options.tls = tls_option;
opts.postgres_options = Some(postgres_options);
}
Ok(opts)
}
}
#[cfg(test)]
mod tests {
use std::time::Duration;
use servers::auth::{Identity, Password, UserProviderRef};
use super::*;
#[test]
@@ -269,12 +294,23 @@ mod tests {
std::env::current_dir().unwrap().as_path().to_str().unwrap()
)),
influxdb_enable: false,
enable_memory_catalog: false,
tls_mode: None,
tls_cert_path: None,
tls_key_path: None,
user_provider: None,
};
let fe_opts = FrontendOptions::try_from(cmd).unwrap();
assert_eq!(Mode::Standalone, fe_opts.mode);
assert_eq!("127.0.0.1:3001".to_string(), fe_opts.datanode_rpc_addr);
assert_eq!(Some("127.0.0.1:4000".to_string()), fe_opts.http_addr);
assert_eq!(
"127.0.0.1:4000".to_string(),
fe_opts.http_options.as_ref().unwrap().addr
);
assert_eq!(
Duration::from_secs(30),
fe_opts.http_options.as_ref().unwrap().timeout
);
assert_eq!(
"127.0.0.1:4001".to_string(),
fe_opts.grpc_options.unwrap().addr
@@ -286,4 +322,33 @@ mod tests {
assert_eq!(2, fe_opts.mysql_options.as_ref().unwrap().runtime_size);
assert!(fe_opts.influxdb_options.as_ref().unwrap().enable);
}
#[tokio::test]
async fn test_try_from_start_command_to_anymap() {
let command = StartCommand {
http_addr: None,
rpc_addr: None,
mysql_addr: None,
postgres_addr: None,
opentsdb_addr: None,
config_file: None,
influxdb_enable: false,
enable_memory_catalog: false,
tls_mode: None,
tls_cert_path: None,
tls_key_path: None,
user_provider: Some("static_user_provider:cmd:test=test".to_string()),
};
let plugins = load_frontend_plugins(&command.user_provider);
assert!(plugins.is_ok());
let plugins = plugins.unwrap();
let provider = plugins.get::<UserProviderRef>();
assert!(provider.is_some());
let provider = provider.unwrap();
let result = provider
.auth(Identity::UserId("test", None), Password::PlainText("test"))
.await;
assert!(result.is_ok());
}
}

View File

@@ -12,8 +12,8 @@
// See the License for the specific language governing permissions and
// limitations under the License.
use bitvec::prelude as bv;
pub use bitvec::prelude;
// `Lsb0` provides the best codegen for bit manipulation,
// see https://github.com/bitvecto-rs/bitvec/blob/main/doc/order/Lsb0.md
pub type BitVec = bv::BitVec<u8, bv::Lsb0>;
pub type BitVec = prelude::BitVec<u8>;

View File

@@ -12,8 +12,8 @@
// See the License for the specific language governing permissions and
// limitations under the License.
pub mod bitset;
pub mod bit_vec;
pub mod buffer;
pub mod bytes;
pub use bitset::BitVec;
pub use bit_vec::BitVec;

View File

@@ -14,7 +14,6 @@ regex = "1.6"
serde = "1.0"
serde_json = "1.0"
snafu = { version = "0.7", features = ["backtraces"] }
table = { path = "../../table" }
[dev-dependencies]
chrono = "0.4"

View File

@@ -25,9 +25,3 @@ pub const MIN_USER_TABLE_ID: u32 = 1024;
pub const SYSTEM_CATALOG_TABLE_ID: u32 = 0;
/// scripts table id
pub const SCRIPTS_TABLE_ID: u32 = 1;
pub(crate) const CATALOG_KEY_PREFIX: &str = "__c";
pub(crate) const SCHEMA_KEY_PREFIX: &str = "__s";
pub(crate) const TABLE_GLOBAL_KEY_PREFIX: &str = "__tg";
pub(crate) const TABLE_REGIONAL_KEY_PREFIX: &str = "__tr";
pub const TABLE_ID_KEY_PREFIX: &str = "__tid";

View File

@@ -14,10 +14,3 @@
pub mod consts;
pub mod error;
mod helper;
pub use helper::{
build_catalog_prefix, build_schema_prefix, build_table_global_prefix,
build_table_regional_prefix, CatalogKey, CatalogValue, SchemaKey, SchemaValue, TableGlobalKey,
TableGlobalValue, TableRegionalKey, TableRegionalValue,
};

View File

@@ -62,6 +62,19 @@ pub enum StatusCode {
/// Runtime resources exhausted, like creating threads failed.
RuntimeResourcesExhausted = 6000,
// ====== End of server related status code =======
// ====== Begin of auth related status code =====
/// User not exist
UserNotFound = 7000,
/// Unsupported password type
UnsupportedPasswordType = 7001,
/// Username and password does not match
UserPasswordMismatch = 7002,
/// Not found http authorization header
AuthHeaderNotFound = 7003,
/// Invalid http authorization header
InvalidAuthHeader = 7004,
// ====== End of auth related status code =====
}
impl StatusCode {

View File

@@ -9,9 +9,9 @@ arc-swap = "1.0"
chrono-tz = "0.6"
common-error = { path = "../error" }
common-function-macro = { path = "../function-macro" }
common-time = { path = "../time" }
common-query = { path = "../query" }
datafusion-common = { git = "https://github.com/apache/arrow-datafusion.git", branch = "arrow2" }
common-time = { path = "../time" }
datafusion-common = "14.0.0"
datatypes = { path = "../../datatypes" }
libc = "0.2"
num = "0.4"
@@ -21,20 +21,6 @@ paste = "1.0"
snafu = { version = "0.7", features = ["backtraces"] }
statrs = "0.15"
[dependencies.arrow]
features = [
"io_csv",
"io_json",
"io_parquet",
"io_parquet_compression",
"io_ipc",
"ahash",
"compute",
"serde_types",
]
package = "arrow2"
version = "0.10"
[dev-dependencies]
ron = "0.7"
serde = { version = "1.0", features = ["derive"] }

View File

@@ -12,5 +12,4 @@
// See the License for the specific language governing permissions and
// limitations under the License.
pub mod error;
pub mod scalars;

View File

@@ -23,6 +23,5 @@ pub(crate) mod test;
mod timestamp;
pub mod udf;
pub use aggregate::MedianAccumulatorCreator;
pub use function::{Function, FunctionRef};
pub use function_registry::{FunctionRegistry, FUNCTION_REGISTRY};

View File

@@ -16,7 +16,6 @@ mod argmax;
mod argmin;
mod diff;
mod mean;
mod median;
mod percentile;
mod polyval;
mod scipy_stats_norm_cdf;
@@ -29,7 +28,6 @@ pub use argmin::ArgminAccumulatorCreator;
use common_query::logical_plan::AggregateFunctionCreatorRef;
pub use diff::DiffAccumulatorCreator;
pub use mean::MeanAccumulatorCreator;
pub use median::MedianAccumulatorCreator;
pub use percentile::PercentileAccumulatorCreator;
pub use polyval::PolyvalAccumulatorCreator;
pub use scipy_stats_norm_cdf::ScipyStatsNormCdfAccumulatorCreator;
@@ -88,7 +86,6 @@ impl AggregateFunctions {
};
}
register_aggr_func!("median", 1, MedianAccumulatorCreator);
register_aggr_func!("diff", 1, DiffAccumulatorCreator);
register_aggr_func!("mean", 1, MeanAccumulatorCreator);
register_aggr_func!("polyval", 2, PolyvalAccumulatorCreator);

View File

@@ -20,24 +20,22 @@ use common_query::error::{BadAccumulatorImplSnafu, CreateAccumulatorSnafu, Resul
use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
use common_query::prelude::*;
use datatypes::prelude::*;
use datatypes::vectors::ConstantVector;
use datatypes::types::{LogicalPrimitiveType, WrapperType};
use datatypes::vectors::{ConstantVector, Helper};
use datatypes::with_match_primitive_type_id;
use snafu::ensure;
// https://numpy.org/doc/stable/reference/generated/numpy.argmax.html
// return the index of the max value
#[derive(Debug, Default)]
pub struct Argmax<T>
where
T: Primitive + PartialOrd,
{
pub struct Argmax<T> {
max: Option<T>,
n: u64,
}
impl<T> Argmax<T>
where
T: Primitive + PartialOrd,
T: PartialOrd + Copy,
{
fn update(&mut self, value: T, index: u64) {
if let Some(Ordering::Less) = self.max.partial_cmp(&Some(value)) {
@@ -49,8 +47,7 @@ where
impl<T> Accumulator for Argmax<T>
where
T: Primitive + PartialOrd,
for<'a> T: Scalar<RefType<'a> = T>,
T: WrapperType + PartialOrd,
{
fn state(&self) -> Result<Vec<Value>> {
match self.max {
@@ -66,10 +63,10 @@ where
let column = &values[0];
let column: &<T as Scalar>::VectorType = if column.is_const() {
let column: &ConstantVector = unsafe { VectorHelper::static_cast(column) };
unsafe { VectorHelper::static_cast(column.inner()) }
let column: &ConstantVector = unsafe { Helper::static_cast(column) };
unsafe { Helper::static_cast(column.inner()) }
} else {
unsafe { VectorHelper::static_cast(column) }
unsafe { Helper::static_cast(column) }
};
for (i, v) in column.iter_data().enumerate() {
if let Some(value) = v {
@@ -93,8 +90,8 @@ where
let max = &states[0];
let index = &states[1];
let max: &<T as Scalar>::VectorType = unsafe { VectorHelper::static_cast(max) };
let index: &<u64 as Scalar>::VectorType = unsafe { VectorHelper::static_cast(index) };
let max: &<T as Scalar>::VectorType = unsafe { Helper::static_cast(max) };
let index: &<u64 as Scalar>::VectorType = unsafe { Helper::static_cast(index) };
index
.iter_data()
.flatten()
@@ -122,7 +119,7 @@ impl AggregateFunctionCreator for ArgmaxAccumulatorCreator {
with_match_primitive_type_id!(
input_type.logical_type_id(),
|$S| {
Ok(Box::new(Argmax::<$S>::default()))
Ok(Box::new(Argmax::<<$S as LogicalPrimitiveType>::Wrapper>::default()))
},
{
let err_msg = format!(
@@ -154,7 +151,7 @@ impl AggregateFunctionCreator for ArgmaxAccumulatorCreator {
#[cfg(test)]
mod test {
use datatypes::vectors::PrimitiveVector;
use datatypes::vectors::Int32Vector;
use super::*;
#[test]
@@ -166,21 +163,19 @@ mod test {
// test update one not-null value
let mut argmax = Argmax::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![Some(42)]))];
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![Some(42)]))];
assert!(argmax.update_batch(&v).is_ok());
assert_eq!(Value::from(0_u64), argmax.evaluate().unwrap());
// test update one null value
let mut argmax = Argmax::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
Option::<i32>::None,
]))];
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![Option::<i32>::None]))];
assert!(argmax.update_batch(&v).is_ok());
assert_eq!(Value::Null, argmax.evaluate().unwrap());
// test update no null-value batch
let mut argmax = Argmax::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![
Some(-1i32),
Some(1),
Some(3),
@@ -190,7 +185,7 @@ mod test {
// test update null-value batch
let mut argmax = Argmax::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![
Some(-2i32),
None,
Some(4),
@@ -201,7 +196,7 @@ mod test {
// test update with constant vector
let mut argmax = Argmax::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(ConstantVector::new(
Arc::new(PrimitiveVector::<i32>::from_vec(vec![4])),
Arc::new(Int32Vector::from_vec(vec![4])),
10,
))];
assert!(argmax.update_batch(&v).is_ok());

View File

@@ -20,23 +20,20 @@ use common_query::error::{BadAccumulatorImplSnafu, CreateAccumulatorSnafu, Resul
use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
use common_query::prelude::*;
use datatypes::prelude::*;
use datatypes::vectors::ConstantVector;
use datatypes::vectors::{ConstantVector, Helper};
use datatypes::with_match_primitive_type_id;
use snafu::ensure;
// // https://numpy.org/doc/stable/reference/generated/numpy.argmin.html
#[derive(Debug, Default)]
pub struct Argmin<T>
where
T: Primitive + PartialOrd,
{
pub struct Argmin<T> {
min: Option<T>,
n: u32,
}
impl<T> Argmin<T>
where
T: Primitive + PartialOrd,
T: Copy + PartialOrd,
{
fn update(&mut self, value: T, index: u32) {
match self.min {
@@ -56,8 +53,7 @@ where
impl<T> Accumulator for Argmin<T>
where
T: Primitive + PartialOrd,
for<'a> T: Scalar<RefType<'a> = T>,
T: WrapperType + PartialOrd,
{
fn state(&self) -> Result<Vec<Value>> {
match self.min {
@@ -75,10 +71,10 @@ where
let column = &values[0];
let column: &<T as Scalar>::VectorType = if column.is_const() {
let column: &ConstantVector = unsafe { VectorHelper::static_cast(column) };
unsafe { VectorHelper::static_cast(column.inner()) }
let column: &ConstantVector = unsafe { Helper::static_cast(column) };
unsafe { Helper::static_cast(column.inner()) }
} else {
unsafe { VectorHelper::static_cast(column) }
unsafe { Helper::static_cast(column) }
};
for (i, v) in column.iter_data().enumerate() {
if let Some(value) = v {
@@ -102,8 +98,8 @@ where
let min = &states[0];
let index = &states[1];
let min: &<T as Scalar>::VectorType = unsafe { VectorHelper::static_cast(min) };
let index: &<u32 as Scalar>::VectorType = unsafe { VectorHelper::static_cast(index) };
let min: &<T as Scalar>::VectorType = unsafe { Helper::static_cast(min) };
let index: &<u32 as Scalar>::VectorType = unsafe { Helper::static_cast(index) };
index
.iter_data()
.flatten()
@@ -131,7 +127,7 @@ impl AggregateFunctionCreator for ArgminAccumulatorCreator {
with_match_primitive_type_id!(
input_type.logical_type_id(),
|$S| {
Ok(Box::new(Argmin::<$S>::default()))
Ok(Box::new(Argmin::<<$S as LogicalPrimitiveType>::Wrapper>::default()))
},
{
let err_msg = format!(
@@ -163,7 +159,7 @@ impl AggregateFunctionCreator for ArgminAccumulatorCreator {
#[cfg(test)]
mod test {
use datatypes::vectors::PrimitiveVector;
use datatypes::vectors::Int32Vector;
use super::*;
#[test]
@@ -175,21 +171,19 @@ mod test {
// test update one not-null value
let mut argmin = Argmin::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![Some(42)]))];
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![Some(42)]))];
assert!(argmin.update_batch(&v).is_ok());
assert_eq!(Value::from(0_u32), argmin.evaluate().unwrap());
// test update one null value
let mut argmin = Argmin::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
Option::<i32>::None,
]))];
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![Option::<i32>::None]))];
assert!(argmin.update_batch(&v).is_ok());
assert_eq!(Value::Null, argmin.evaluate().unwrap());
// test update no null-value batch
let mut argmin = Argmin::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![
Some(-1i32),
Some(1),
Some(3),
@@ -199,7 +193,7 @@ mod test {
// test update null-value batch
let mut argmin = Argmin::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![
Some(-2i32),
None,
Some(4),
@@ -210,7 +204,7 @@ mod test {
// test update with constant vector
let mut argmin = Argmin::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(ConstantVector::new(
Arc::new(PrimitiveVector::<i32>::from_vec(vec![4])),
Arc::new(Int32Vector::from_vec(vec![4])),
10,
))];
assert!(argmin.update_batch(&v).is_ok());

View File

@@ -22,40 +22,32 @@ use common_query::error::{
use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
use common_query::prelude::*;
use datatypes::prelude::*;
use datatypes::types::PrimitiveType;
use datatypes::value::ListValue;
use datatypes::vectors::{ConstantVector, ListVector};
use datatypes::vectors::{ConstantVector, Helper, ListVector};
use datatypes::with_match_primitive_type_id;
use num_traits::AsPrimitive;
use snafu::{ensure, OptionExt, ResultExt};
// https://numpy.org/doc/stable/reference/generated/numpy.diff.html
// I is the input type, O is the output type.
#[derive(Debug, Default)]
pub struct Diff<T, SubT>
where
T: Primitive + AsPrimitive<SubT>,
SubT: Primitive + std::ops::Sub<Output = SubT>,
{
values: Vec<T>,
_phantom: PhantomData<SubT>,
pub struct Diff<I, O> {
values: Vec<I>,
_phantom: PhantomData<O>,
}
impl<T, SubT> Diff<T, SubT>
where
T: Primitive + AsPrimitive<SubT>,
SubT: Primitive + std::ops::Sub<Output = SubT>,
{
fn push(&mut self, value: T) {
impl<I, O> Diff<I, O> {
fn push(&mut self, value: I) {
self.values.push(value);
}
}
impl<T, SubT> Accumulator for Diff<T, SubT>
impl<I, O> Accumulator for Diff<I, O>
where
T: Primitive + AsPrimitive<SubT>,
for<'a> T: Scalar<RefType<'a> = T>,
SubT: Primitive + std::ops::Sub<Output = SubT>,
for<'a> SubT: Scalar<RefType<'a> = SubT>,
I: WrapperType,
O: WrapperType,
I::Native: AsPrimitive<O::Native>,
O::Native: std::ops::Sub<Output = O::Native>,
{
fn state(&self) -> Result<Vec<Value>> {
let nums = self
@@ -65,7 +57,7 @@ where
.collect::<Vec<Value>>();
Ok(vec![Value::List(ListValue::new(
Some(Box::new(nums)),
T::default().into().data_type(),
I::LogicalType::build_data_type(),
))])
}
@@ -78,12 +70,12 @@ where
let column = &values[0];
let mut len = 1;
let column: &<T as Scalar>::VectorType = if column.is_const() {
let column: &<I as Scalar>::VectorType = if column.is_const() {
len = column.len();
let column: &ConstantVector = unsafe { VectorHelper::static_cast(column) };
unsafe { VectorHelper::static_cast(column.inner()) }
let column: &ConstantVector = unsafe { Helper::static_cast(column) };
unsafe { Helper::static_cast(column.inner()) }
} else {
unsafe { VectorHelper::static_cast(column) }
unsafe { Helper::static_cast(column) }
};
(0..len).for_each(|_| {
for v in column.iter_data().flatten() {
@@ -109,8 +101,9 @@ where
),
})?;
for state in states.values_iter() {
let state = state.context(FromScalarValueSnafu)?;
self.update_batch(&[state])?
if let Some(state) = state.context(FromScalarValueSnafu)? {
self.update_batch(&[state])?;
}
}
Ok(())
}
@@ -122,11 +115,14 @@ where
let diff = self
.values
.windows(2)
.map(|x| (x[1].as_() - x[0].as_()).into())
.map(|x| {
let native = x[1].into_native().as_() - x[0].into_native().as_();
O::from_native(native).into()
})
.collect::<Vec<Value>>();
let diff = Value::List(ListValue::new(
Some(Box::new(diff)),
SubT::default().into().data_type(),
O::LogicalType::build_data_type(),
));
Ok(diff)
}
@@ -143,7 +139,7 @@ impl AggregateFunctionCreator for DiffAccumulatorCreator {
with_match_primitive_type_id!(
input_type.logical_type_id(),
|$S| {
Ok(Box::new(Diff::<$S,<$S as Primitive>::LargestType>::default()))
Ok(Box::new(Diff::<<$S as LogicalPrimitiveType>::Wrapper, <<$S as LogicalPrimitiveType>::LargestType as LogicalPrimitiveType>::Wrapper>::default()))
},
{
let err_msg = format!(
@@ -163,7 +159,7 @@ impl AggregateFunctionCreator for DiffAccumulatorCreator {
with_match_primitive_type_id!(
input_types[0].logical_type_id(),
|$S| {
Ok(ConcreteDataType::list_datatype(PrimitiveType::<<$S as Primitive>::LargestType>::default().into()))
Ok(ConcreteDataType::list_datatype($S::default().into()))
},
{
unreachable!()
@@ -177,7 +173,7 @@ impl AggregateFunctionCreator for DiffAccumulatorCreator {
with_match_primitive_type_id!(
input_types[0].logical_type_id(),
|$S| {
Ok(vec![ConcreteDataType::list_datatype(PrimitiveType::<$S>::default().into())])
Ok(vec![ConcreteDataType::list_datatype($S::default().into())])
},
{
unreachable!()
@@ -188,9 +184,10 @@ impl AggregateFunctionCreator for DiffAccumulatorCreator {
#[cfg(test)]
mod test {
use datatypes::vectors::PrimitiveVector;
use datatypes::vectors::Int32Vector;
use super::*;
#[test]
fn test_update_batch() {
// test update empty batch, expect not updating anything
@@ -201,21 +198,19 @@ mod test {
// test update one not-null value
let mut diff = Diff::<i32, i64>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![Some(42)]))];
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![Some(42)]))];
assert!(diff.update_batch(&v).is_ok());
assert_eq!(Value::Null, diff.evaluate().unwrap());
// test update one null value
let mut diff = Diff::<i32, i64>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
Option::<i32>::None,
]))];
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![Option::<i32>::None]))];
assert!(diff.update_batch(&v).is_ok());
assert_eq!(Value::Null, diff.evaluate().unwrap());
// test update no null-value batch
let mut diff = Diff::<i32, i64>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![
Some(-1i32),
Some(1),
Some(2),
@@ -232,7 +227,7 @@ mod test {
// test update null-value batch
let mut diff = Diff::<i32, i64>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![
Some(-2i32),
None,
Some(3),
@@ -251,7 +246,7 @@ mod test {
// test update with constant vector
let mut diff = Diff::<i32, i64>::default();
let v: Vec<VectorRef> = vec![Arc::new(ConstantVector::new(
Arc::new(PrimitiveVector::<i32>::from_vec(vec![4])),
Arc::new(Int32Vector::from_vec(vec![4])),
4,
))];
let values = vec![Value::from(0_i64), Value::from(0_i64), Value::from(0_i64)];

View File

@@ -22,16 +22,14 @@ use common_query::error::{
use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
use common_query::prelude::*;
use datatypes::prelude::*;
use datatypes::vectors::{ConstantVector, Float64Vector, UInt64Vector};
use datatypes::types::WrapperType;
use datatypes::vectors::{ConstantVector, Float64Vector, Helper, UInt64Vector};
use datatypes::with_match_primitive_type_id;
use num_traits::AsPrimitive;
use snafu::{ensure, OptionExt};
#[derive(Debug, Default)]
pub struct Mean<T>
where
T: Primitive + AsPrimitive<f64>,
{
pub struct Mean<T> {
sum: f64,
n: u64,
_phantom: PhantomData<T>,
@@ -39,11 +37,12 @@ where
impl<T> Mean<T>
where
T: Primitive + AsPrimitive<f64>,
T: WrapperType,
T::Native: AsPrimitive<f64>,
{
#[inline(always)]
fn push(&mut self, value: T) {
self.sum += value.as_();
self.sum += value.into_native().as_();
self.n += 1;
}
@@ -56,8 +55,8 @@ where
impl<T> Accumulator for Mean<T>
where
T: Primitive + AsPrimitive<f64>,
for<'a> T: Scalar<RefType<'a> = T>,
T: WrapperType,
T::Native: AsPrimitive<f64>,
{
fn state(&self) -> Result<Vec<Value>> {
Ok(vec![self.sum.into(), self.n.into()])
@@ -73,10 +72,10 @@ where
let mut len = 1;
let column: &<T as Scalar>::VectorType = if column.is_const() {
len = column.len();
let column: &ConstantVector = unsafe { VectorHelper::static_cast(column) };
unsafe { VectorHelper::static_cast(column.inner()) }
let column: &ConstantVector = unsafe { Helper::static_cast(column) };
unsafe { Helper::static_cast(column.inner()) }
} else {
unsafe { VectorHelper::static_cast(column) }
unsafe { Helper::static_cast(column) }
};
(0..len).for_each(|_| {
for v in column.iter_data().flatten() {
@@ -150,7 +149,7 @@ impl AggregateFunctionCreator for MeanAccumulatorCreator {
with_match_primitive_type_id!(
input_type.logical_type_id(),
|$S| {
Ok(Box::new(Mean::<$S>::default()))
Ok(Box::new(Mean::<<$S as LogicalPrimitiveType>::Native>::default()))
},
{
let err_msg = format!(
@@ -182,7 +181,7 @@ impl AggregateFunctionCreator for MeanAccumulatorCreator {
#[cfg(test)]
mod test {
use datatypes::vectors::PrimitiveVector;
use datatypes::vectors::Int32Vector;
use super::*;
#[test]
@@ -194,21 +193,19 @@ mod test {
// test update one not-null value
let mut mean = Mean::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![Some(42)]))];
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![Some(42)]))];
assert!(mean.update_batch(&v).is_ok());
assert_eq!(Value::from(42.0_f64), mean.evaluate().unwrap());
// test update one null value
let mut mean = Mean::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
Option::<i32>::None,
]))];
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![Option::<i32>::None]))];
assert!(mean.update_batch(&v).is_ok());
assert_eq!(Value::Null, mean.evaluate().unwrap());
// test update no null-value batch
let mut mean = Mean::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![
Some(-1i32),
Some(1),
Some(2),
@@ -218,7 +215,7 @@ mod test {
// test update null-value batch
let mut mean = Mean::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
let v: Vec<VectorRef> = vec![Arc::new(Int32Vector::from(vec![
Some(-2i32),
None,
Some(3),
@@ -230,7 +227,7 @@ mod test {
// test update with constant vector
let mut mean = Mean::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(ConstantVector::new(
Arc::new(PrimitiveVector::<i32>::from_vec(vec![4])),
Arc::new(Int32Vector::from_vec(vec![4])),
10,
))];
assert!(mean.update_batch(&v).is_ok());

View File

@@ -1,289 +0,0 @@
// Copyright 2022 Greptime Team
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
use std::cmp::Reverse;
use std::collections::BinaryHeap;
use std::sync::Arc;
use common_function_macro::{as_aggr_func_creator, AggrFuncTypeStore};
use common_query::error::{
CreateAccumulatorSnafu, DowncastVectorSnafu, FromScalarValueSnafu, Result,
};
use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
use common_query::prelude::*;
use datatypes::prelude::*;
use datatypes::types::OrdPrimitive;
use datatypes::value::ListValue;
use datatypes::vectors::{ConstantVector, ListVector};
use datatypes::with_match_primitive_type_id;
use num::NumCast;
use snafu::{ensure, OptionExt, ResultExt};
// This median calculation algorithm's details can be found at
// https://leetcode.cn/problems/find-median-from-data-stream/
//
// Basically, it uses two heaps, a maximum heap and a minimum. The maximum heap stores numbers that
// are not greater than the median, and the minimum heap stores the greater. In a streaming of
// numbers, when a number is arrived, we adjust the heaps' tops, so that either one top is the
// median or both tops can be averaged to get the median.
//
// The time complexity to update the median is O(logn), O(1) to get the median; and the space
// complexity is O(n). (Ignore the costs for heap expansion.)
//
// From the point of algorithm, [quick select](https://en.wikipedia.org/wiki/Quickselect) might be
// better. But to use quick select here, we need a mutable self in the final calculation(`evaluate`)
// to swap stored numbers in the states vector. Though we can make our `evaluate` received
// `&mut self`, DataFusion calls our accumulator with `&self` (see `DfAccumulatorAdaptor`). That
// means we have to introduce some kinds of interior mutability, and the overhead is not neglectable.
//
// TODO(LFC): Use quick select to get median when we can modify DataFusion's code, and benchmark with two-heap algorithm.
#[derive(Debug, Default)]
pub struct Median<T>
where
T: Primitive,
{
greater: BinaryHeap<Reverse<OrdPrimitive<T>>>,
not_greater: BinaryHeap<OrdPrimitive<T>>,
}
impl<T> Median<T>
where
T: Primitive,
{
fn push(&mut self, value: T) {
let value = OrdPrimitive::<T>(value);
if self.not_greater.is_empty() {
self.not_greater.push(value);
return;
}
// The `unwrap`s below are safe because there are `push`s before them.
if value <= *self.not_greater.peek().unwrap() {
self.not_greater.push(value);
if self.not_greater.len() > self.greater.len() + 1 {
self.greater.push(Reverse(self.not_greater.pop().unwrap()));
}
} else {
self.greater.push(Reverse(value));
if self.greater.len() > self.not_greater.len() {
self.not_greater.push(self.greater.pop().unwrap().0);
}
}
}
}
// UDAFs are built using the trait `Accumulator`, that offers DataFusion the necessary functions
// to use them.
impl<T> Accumulator for Median<T>
where
T: Primitive,
for<'a> T: Scalar<RefType<'a> = T>,
{
// This function serializes our state to `ScalarValue`, which DataFusion uses to pass this
// state between execution stages. Note that this can be arbitrary data.
//
// The `ScalarValue`s returned here will be passed in as argument `states: &[VectorRef]` to
// `merge_batch` function.
fn state(&self) -> Result<Vec<Value>> {
let nums = self
.greater
.iter()
.map(|x| &x.0)
.chain(self.not_greater.iter())
.map(|&n| n.into())
.collect::<Vec<Value>>();
Ok(vec![Value::List(ListValue::new(
Some(Box::new(nums)),
T::default().into().data_type(),
))])
}
// DataFusion calls this function to update the accumulator's state for a batch of inputs rows.
// It is expected this function to update the accumulator's state.
fn update_batch(&mut self, values: &[VectorRef]) -> Result<()> {
if values.is_empty() {
return Ok(());
}
ensure!(values.len() == 1, InvalidInputStateSnafu);
// This is a unary accumulator, so only one column is provided.
let column = &values[0];
let mut len = 1;
let column: &<T as Scalar>::VectorType = if column.is_const() {
len = column.len();
let column: &ConstantVector = unsafe { VectorHelper::static_cast(column) };
unsafe { VectorHelper::static_cast(column.inner()) }
} else {
unsafe { VectorHelper::static_cast(column) }
};
(0..len).for_each(|_| {
for v in column.iter_data().flatten() {
self.push(v);
}
});
Ok(())
}
// DataFusion executes accumulators in partitions. In some execution stage, DataFusion will
// merge states from other accumulators (returned by `state()` method).
fn merge_batch(&mut self, states: &[VectorRef]) -> Result<()> {
if states.is_empty() {
return Ok(());
}
// The states here are returned by the `state` method. Since we only returned a vector
// with one value in that method, `states[0]` is fine.
let states = &states[0];
let states = states
.as_any()
.downcast_ref::<ListVector>()
.with_context(|| DowncastVectorSnafu {
err_msg: format!(
"expect ListVector, got vector type {}",
states.vector_type_name()
),
})?;
for state in states.values_iter() {
let state = state.context(FromScalarValueSnafu)?;
// merging state is simply accumulate stored numbers from others', so just call update
self.update_batch(&[state])?
}
Ok(())
}
// DataFusion expects this function to return the final value of this aggregator.
fn evaluate(&self) -> Result<Value> {
if self.not_greater.is_empty() {
assert!(
self.greater.is_empty(),
"not expected in two-heap median algorithm, there must be a bug when implementing it"
);
return Ok(Value::Null);
}
// unwrap is safe because we checked not_greater heap's len above
let not_greater = *self.not_greater.peek().unwrap();
let median = if self.not_greater.len() > self.greater.len() {
not_greater.into()
} else {
// unwrap is safe because greater heap len >= not_greater heap len, which is > 0
let greater = self.greater.peek().unwrap();
// the following three NumCast's `unwrap`s are safe because T is primitive
let not_greater_v: f64 = NumCast::from(not_greater.as_primitive()).unwrap();
let greater_v: f64 = NumCast::from(greater.0.as_primitive()).unwrap();
let median: T = NumCast::from((not_greater_v + greater_v) / 2.0).unwrap();
median.into()
};
Ok(median)
}
}
#[as_aggr_func_creator]
#[derive(Debug, Default, AggrFuncTypeStore)]
pub struct MedianAccumulatorCreator {}
impl AggregateFunctionCreator for MedianAccumulatorCreator {
fn creator(&self) -> AccumulatorCreatorFunction {
let creator: AccumulatorCreatorFunction = Arc::new(move |types: &[ConcreteDataType]| {
let input_type = &types[0];
with_match_primitive_type_id!(
input_type.logical_type_id(),
|$S| {
Ok(Box::new(Median::<$S>::default()))
},
{
let err_msg = format!(
"\"MEDIAN\" aggregate function not support data type {:?}",
input_type.logical_type_id(),
);
CreateAccumulatorSnafu { err_msg }.fail()?
}
)
});
creator
}
fn output_type(&self) -> Result<ConcreteDataType> {
let input_types = self.input_types()?;
ensure!(input_types.len() == 1, InvalidInputStateSnafu);
// unwrap is safe because we have checked input_types len must equals 1
Ok(input_types.into_iter().next().unwrap())
}
fn state_types(&self) -> Result<Vec<ConcreteDataType>> {
Ok(vec![ConcreteDataType::list_datatype(self.output_type()?)])
}
}
#[cfg(test)]
mod test {
use datatypes::vectors::PrimitiveVector;
use super::*;
#[test]
fn test_update_batch() {
// test update empty batch, expect not updating anything
let mut median = Median::<i32>::default();
assert!(median.update_batch(&[]).is_ok());
assert!(median.not_greater.is_empty());
assert!(median.greater.is_empty());
assert_eq!(Value::Null, median.evaluate().unwrap());
// test update one not-null value
let mut median = Median::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![Some(42)]))];
assert!(median.update_batch(&v).is_ok());
assert_eq!(Value::Int32(42), median.evaluate().unwrap());
// test update one null value
let mut median = Median::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
Option::<i32>::None,
]))];
assert!(median.update_batch(&v).is_ok());
assert_eq!(Value::Null, median.evaluate().unwrap());
// test update no null-value batch
let mut median = Median::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
Some(-1i32),
Some(1),
Some(2),
]))];
assert!(median.update_batch(&v).is_ok());
assert_eq!(Value::Int32(1), median.evaluate().unwrap());
// test update null-value batch
let mut median = Median::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(PrimitiveVector::<i32>::from(vec![
Some(-2i32),
None,
Some(3),
Some(4),
]))];
assert!(median.update_batch(&v).is_ok());
assert_eq!(Value::Int32(3), median.evaluate().unwrap());
// test update with constant vector
let mut median = Median::<i32>::default();
let v: Vec<VectorRef> = vec![Arc::new(ConstantVector::new(
Arc::new(PrimitiveVector::<i32>::from_vec(vec![4])),
10,
))];
assert!(median.update_batch(&v).is_ok());
assert_eq!(Value::Int32(4), median.evaluate().unwrap());
}
}

View File

@@ -26,7 +26,7 @@ use common_query::prelude::*;
use datatypes::prelude::*;
use datatypes::types::OrdPrimitive;
use datatypes::value::{ListValue, OrderedFloat};
use datatypes::vectors::{ConstantVector, Float64Vector, ListVector};
use datatypes::vectors::{ConstantVector, Float64Vector, Helper, ListVector};
use datatypes::with_match_primitive_type_id;
use num::NumCast;
use snafu::{ensure, OptionExt, ResultExt};
@@ -44,15 +44,15 @@ use snafu::{ensure, OptionExt, ResultExt};
// This optional method parameter specifies the method to use when the desired quantile lies between two data points i < j.
// If g is the fractional part of the index surrounded by i and alpha and beta are correction constants modifying i and j.
// i+g = (q-alpha)/(n-alpha-beta+1)
// Below, q is the quantile value, n is the sample size and alpha and beta are constants. The following formula gives an interpolation i + g of where the quantile would be in the sorted sample.
// With i being the floor and g the fractional part of the result.
// Below, 'q' is the quantile value, 'n' is the sample size and alpha and beta are constants. The following formula gives an interpolation "i + g" of where the quantile would be in the sorted sample.
// With 'i' being the floor and 'g' the fractional part of the result.
// the default method is linear where
// alpha = 1
// beta = 1
#[derive(Debug, Default)]
pub struct Percentile<T>
where
T: Primitive,
T: WrapperType,
{
greater: BinaryHeap<Reverse<OrdPrimitive<T>>>,
not_greater: BinaryHeap<OrdPrimitive<T>>,
@@ -62,7 +62,7 @@ where
impl<T> Percentile<T>
where
T: Primitive,
T: WrapperType,
{
fn push(&mut self, value: T) {
let value = OrdPrimitive::<T>(value);
@@ -93,8 +93,7 @@ where
impl<T> Accumulator for Percentile<T>
where
T: Primitive,
for<'a> T: Scalar<RefType<'a> = T>,
T: WrapperType,
{
fn state(&self) -> Result<Vec<Value>> {
let nums = self
@@ -107,7 +106,7 @@ where
Ok(vec![
Value::List(ListValue::new(
Some(Box::new(nums)),
T::default().into().data_type(),
T::LogicalType::build_data_type(),
)),
self.p.into(),
])
@@ -129,14 +128,14 @@ where
let mut len = 1;
let column: &<T as Scalar>::VectorType = if column.is_const() {
len = column.len();
let column: &ConstantVector = unsafe { VectorHelper::static_cast(column) };
unsafe { VectorHelper::static_cast(column.inner()) }
let column: &ConstantVector = unsafe { Helper::static_cast(column) };
unsafe { Helper::static_cast(column.inner()) }
} else {
unsafe { VectorHelper::static_cast(column) }
unsafe { Helper::static_cast(column) }
};
let x = &values[1];
let x = VectorHelper::check_get_scalar::<f64>(x).context(error::InvalidInputsSnafu {
let x = Helper::check_get_scalar::<f64>(x).context(error::InvalidInputTypeSnafu {
err_msg: "expecting \"POLYVAL\" function's second argument to be float64",
})?;
// `get(0)` is safe because we have checked `values[1].len() == values[0].len() != 0`
@@ -209,10 +208,11 @@ where
),
})?;
for value in values.values_iter() {
let value = value.context(FromScalarValueSnafu)?;
let column: &<T as Scalar>::VectorType = unsafe { VectorHelper::static_cast(&value) };
for v in column.iter_data().flatten() {
self.push(v);
if let Some(value) = value.context(FromScalarValueSnafu)? {
let column: &<T as Scalar>::VectorType = unsafe { Helper::static_cast(&value) };
for v in column.iter_data().flatten() {
self.push(v);
}
}
}
Ok(())
@@ -259,7 +259,7 @@ impl AggregateFunctionCreator for PercentileAccumulatorCreator {
with_match_primitive_type_id!(
input_type.logical_type_id(),
|$S| {
Ok(Box::new(Percentile::<$S>::default()))
Ok(Box::new(Percentile::<<$S as LogicalPrimitiveType>::Wrapper>::default()))
},
{
let err_msg = format!(
@@ -292,7 +292,7 @@ impl AggregateFunctionCreator for PercentileAccumulatorCreator {
#[cfg(test)]
mod test {
use datatypes::vectors::PrimitiveVector;
use datatypes::vectors::{Float64Vector, Int32Vector};
use super::*;
#[test]
@@ -307,8 +307,8 @@ mod test {
// test update one not-null value
let mut percentile = Percentile::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![Some(42)])),
Arc::new(PrimitiveVector::<f64>::from(vec![Some(100.0_f64)])),
Arc::new(Int32Vector::from(vec![Some(42)])),
Arc::new(Float64Vector::from(vec![Some(100.0_f64)])),
];
assert!(percentile.update_batch(&v).is_ok());
assert_eq!(Value::from(42.0_f64), percentile.evaluate().unwrap());
@@ -316,8 +316,8 @@ mod test {
// test update one null value
let mut percentile = Percentile::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![Option::<i32>::None])),
Arc::new(PrimitiveVector::<f64>::from(vec![Some(100.0_f64)])),
Arc::new(Int32Vector::from(vec![Option::<i32>::None])),
Arc::new(Float64Vector::from(vec![Some(100.0_f64)])),
];
assert!(percentile.update_batch(&v).is_ok());
assert_eq!(Value::Null, percentile.evaluate().unwrap());
@@ -325,12 +325,8 @@ mod test {
// test update no null-value batch
let mut percentile = Percentile::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![
Some(-1i32),
Some(1),
Some(2),
])),
Arc::new(PrimitiveVector::<f64>::from(vec![
Arc::new(Int32Vector::from(vec![Some(-1i32), Some(1), Some(2)])),
Arc::new(Float64Vector::from(vec![
Some(100.0_f64),
Some(100.0_f64),
Some(100.0_f64),
@@ -342,13 +338,8 @@ mod test {
// test update null-value batch
let mut percentile = Percentile::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![
Some(-2i32),
None,
Some(3),
Some(4),
])),
Arc::new(PrimitiveVector::<f64>::from(vec![
Arc::new(Int32Vector::from(vec![Some(-2i32), None, Some(3), Some(4)])),
Arc::new(Float64Vector::from(vec![
Some(100.0_f64),
Some(100.0_f64),
Some(100.0_f64),
@@ -362,13 +353,10 @@ mod test {
let mut percentile = Percentile::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(ConstantVector::new(
Arc::new(PrimitiveVector::<i32>::from_vec(vec![4])),
Arc::new(Int32Vector::from_vec(vec![4])),
2,
)),
Arc::new(PrimitiveVector::<f64>::from(vec![
Some(100.0_f64),
Some(100.0_f64),
])),
Arc::new(Float64Vector::from(vec![Some(100.0_f64), Some(100.0_f64)])),
];
assert!(percentile.update_batch(&v).is_ok());
assert_eq!(Value::from(4_f64), percentile.evaluate().unwrap());
@@ -376,12 +364,8 @@ mod test {
// test left border
let mut percentile = Percentile::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![
Some(-1i32),
Some(1),
Some(2),
])),
Arc::new(PrimitiveVector::<f64>::from(vec![
Arc::new(Int32Vector::from(vec![Some(-1i32), Some(1), Some(2)])),
Arc::new(Float64Vector::from(vec![
Some(0.0_f64),
Some(0.0_f64),
Some(0.0_f64),
@@ -393,12 +377,8 @@ mod test {
// test medium
let mut percentile = Percentile::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![
Some(-1i32),
Some(1),
Some(2),
])),
Arc::new(PrimitiveVector::<f64>::from(vec![
Arc::new(Int32Vector::from(vec![Some(-1i32), Some(1), Some(2)])),
Arc::new(Float64Vector::from(vec![
Some(50.0_f64),
Some(50.0_f64),
Some(50.0_f64),
@@ -410,12 +390,8 @@ mod test {
// test right border
let mut percentile = Percentile::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![
Some(-1i32),
Some(1),
Some(2),
])),
Arc::new(PrimitiveVector::<f64>::from(vec![
Arc::new(Int32Vector::from(vec![Some(-1i32), Some(1), Some(2)])),
Arc::new(Float64Vector::from(vec![
Some(100.0_f64),
Some(100.0_f64),
Some(100.0_f64),
@@ -431,12 +407,8 @@ mod test {
// >> 6.400000000000
let mut percentile = Percentile::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![
Some(10i32),
Some(7),
Some(4),
])),
Arc::new(PrimitiveVector::<f64>::from(vec![
Arc::new(Int32Vector::from(vec![Some(10i32), Some(7), Some(4)])),
Arc::new(Float64Vector::from(vec![
Some(40.0_f64),
Some(40.0_f64),
Some(40.0_f64),
@@ -451,12 +423,8 @@ mod test {
// >> 9.7000000000000011
let mut percentile = Percentile::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![
Some(10i32),
Some(7),
Some(4),
])),
Arc::new(PrimitiveVector::<f64>::from(vec![
Arc::new(Int32Vector::from(vec![Some(10i32), Some(7), Some(4)])),
Arc::new(Float64Vector::from(vec![
Some(95.0_f64),
Some(95.0_f64),
Some(95.0_f64),

View File

@@ -23,9 +23,9 @@ use common_query::error::{
use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
use common_query::prelude::*;
use datatypes::prelude::*;
use datatypes::types::PrimitiveType;
use datatypes::types::{LogicalPrimitiveType, WrapperType};
use datatypes::value::ListValue;
use datatypes::vectors::{ConstantVector, Int64Vector, ListVector};
use datatypes::vectors::{ConstantVector, Helper, Int64Vector, ListVector};
use datatypes::with_match_primitive_type_id;
use num_traits::AsPrimitive;
use snafu::{ensure, OptionExt, ResultExt};
@@ -34,8 +34,10 @@ use snafu::{ensure, OptionExt, ResultExt};
#[derive(Debug, Default)]
pub struct Polyval<T, PolyT>
where
T: Primitive + AsPrimitive<PolyT>,
PolyT: Primitive + std::ops::Mul<Output = PolyT>,
T: WrapperType,
T::Native: AsPrimitive<PolyT::Native>,
PolyT: WrapperType,
PolyT::Native: std::ops::Mul<Output = PolyT::Native>,
{
values: Vec<T>,
// DataFusion casts constant in into i64 type.
@@ -45,8 +47,10 @@ where
impl<T, PolyT> Polyval<T, PolyT>
where
T: Primitive + AsPrimitive<PolyT>,
PolyT: Primitive + std::ops::Mul<Output = PolyT>,
T: WrapperType,
T::Native: AsPrimitive<PolyT::Native>,
PolyT: WrapperType,
PolyT::Native: std::ops::Mul<Output = PolyT::Native>,
{
fn push(&mut self, value: T) {
self.values.push(value);
@@ -55,11 +59,11 @@ where
impl<T, PolyT> Accumulator for Polyval<T, PolyT>
where
T: Primitive + AsPrimitive<PolyT>,
PolyT: Primitive + std::ops::Mul<Output = PolyT> + std::iter::Sum<PolyT>,
for<'a> T: Scalar<RefType<'a> = T>,
for<'a> PolyT: Scalar<RefType<'a> = PolyT>,
i64: AsPrimitive<PolyT>,
T: WrapperType,
T::Native: AsPrimitive<PolyT::Native>,
PolyT: WrapperType + std::iter::Sum<<PolyT as WrapperType>::Native>,
PolyT::Native: std::ops::Mul<Output = PolyT::Native> + std::iter::Sum<PolyT::Native>,
i64: AsPrimitive<<PolyT as WrapperType>::Native>,
{
fn state(&self) -> Result<Vec<Value>> {
let nums = self
@@ -70,7 +74,7 @@ where
Ok(vec![
Value::List(ListValue::new(
Some(Box::new(nums)),
T::default().into().data_type(),
T::LogicalType::build_data_type(),
)),
self.x.into(),
])
@@ -91,10 +95,10 @@ where
let mut len = 1;
let column: &<T as Scalar>::VectorType = if column.is_const() {
len = column.len();
let column: &ConstantVector = unsafe { VectorHelper::static_cast(column) };
unsafe { VectorHelper::static_cast(column.inner()) }
let column: &ConstantVector = unsafe { Helper::static_cast(column) };
unsafe { Helper::static_cast(column.inner()) }
} else {
unsafe { VectorHelper::static_cast(column) }
unsafe { Helper::static_cast(column) }
};
(0..len).for_each(|_| {
for v in column.iter_data().flatten() {
@@ -103,7 +107,7 @@ where
});
let x = &values[1];
let x = VectorHelper::check_get_scalar::<i64>(x).context(error::InvalidInputsSnafu {
let x = Helper::check_get_scalar::<i64>(x).context(error::InvalidInputTypeSnafu {
err_msg: "expecting \"POLYVAL\" function's second argument to be a positive integer",
})?;
// `get(0)` is safe because we have checked `values[1].len() == values[0].len() != 0`
@@ -172,12 +176,14 @@ where
),
})?;
for value in values.values_iter() {
let value = value.context(FromScalarValueSnafu)?;
let column: &<T as Scalar>::VectorType = unsafe { VectorHelper::static_cast(&value) };
for v in column.iter_data().flatten() {
self.push(v);
if let Some(value) = value.context(FromScalarValueSnafu)? {
let column: &<T as Scalar>::VectorType = unsafe { Helper::static_cast(&value) };
for v in column.iter_data().flatten() {
self.push(v);
}
}
}
Ok(())
}
@@ -196,7 +202,7 @@ where
.values
.iter()
.enumerate()
.map(|(i, &value)| value.as_() * (x.pow((len - 1 - i) as u32)).as_())
.map(|(i, &value)| value.into_native().as_() * x.pow((len - 1 - i) as u32).as_())
.sum();
Ok(polyval.into())
}
@@ -213,7 +219,7 @@ impl AggregateFunctionCreator for PolyvalAccumulatorCreator {
with_match_primitive_type_id!(
input_type.logical_type_id(),
|$S| {
Ok(Box::new(Polyval::<$S,<$S as Primitive>::LargestType>::default()))
Ok(Box::new(Polyval::<<$S as LogicalPrimitiveType>::Wrapper, <<$S as LogicalPrimitiveType>::LargestType as LogicalPrimitiveType>::Wrapper>::default()))
},
{
let err_msg = format!(
@@ -234,7 +240,7 @@ impl AggregateFunctionCreator for PolyvalAccumulatorCreator {
with_match_primitive_type_id!(
input_type,
|$S| {
Ok(PrimitiveType::<<$S as Primitive>::LargestType>::default().into())
Ok(<<$S as LogicalPrimitiveType>::LargestType as LogicalPrimitiveType>::build_data_type())
},
{
unreachable!()
@@ -254,7 +260,7 @@ impl AggregateFunctionCreator for PolyvalAccumulatorCreator {
#[cfg(test)]
mod test {
use datatypes::vectors::PrimitiveVector;
use datatypes::vectors::Int32Vector;
use super::*;
#[test]
@@ -268,8 +274,8 @@ mod test {
// test update one not-null value
let mut polyval = Polyval::<i32, i64>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![Some(3)])),
Arc::new(PrimitiveVector::<i64>::from(vec![Some(2_i64)])),
Arc::new(Int32Vector::from(vec![Some(3)])),
Arc::new(Int64Vector::from(vec![Some(2_i64)])),
];
assert!(polyval.update_batch(&v).is_ok());
assert_eq!(Value::Int64(3), polyval.evaluate().unwrap());
@@ -277,8 +283,8 @@ mod test {
// test update one null value
let mut polyval = Polyval::<i32, i64>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![Option::<i32>::None])),
Arc::new(PrimitiveVector::<i64>::from(vec![Some(2_i64)])),
Arc::new(Int32Vector::from(vec![Option::<i32>::None])),
Arc::new(Int64Vector::from(vec![Some(2_i64)])),
];
assert!(polyval.update_batch(&v).is_ok());
assert_eq!(Value::Null, polyval.evaluate().unwrap());
@@ -286,12 +292,8 @@ mod test {
// test update no null-value batch
let mut polyval = Polyval::<i32, i64>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![
Some(3),
Some(0),
Some(1),
])),
Arc::new(PrimitiveVector::<i64>::from(vec![
Arc::new(Int32Vector::from(vec![Some(3), Some(0), Some(1)])),
Arc::new(Int64Vector::from(vec![
Some(2_i64),
Some(2_i64),
Some(2_i64),
@@ -303,13 +305,8 @@ mod test {
// test update null-value batch
let mut polyval = Polyval::<i32, i64>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![
Some(3),
Some(0),
None,
Some(1),
])),
Arc::new(PrimitiveVector::<i64>::from(vec![
Arc::new(Int32Vector::from(vec![Some(3), Some(0), None, Some(1)])),
Arc::new(Int64Vector::from(vec![
Some(2_i64),
Some(2_i64),
Some(2_i64),
@@ -323,10 +320,10 @@ mod test {
let mut polyval = Polyval::<i32, i64>::default();
let v: Vec<VectorRef> = vec![
Arc::new(ConstantVector::new(
Arc::new(PrimitiveVector::<i32>::from_vec(vec![4])),
Arc::new(Int32Vector::from_vec(vec![4])),
2,
)),
Arc::new(PrimitiveVector::<i64>::from(vec![Some(5_i64), Some(5_i64)])),
Arc::new(Int64Vector::from(vec![Some(5_i64), Some(5_i64)])),
];
assert!(polyval.update_batch(&v).is_ok());
assert_eq!(Value::Int64(24), polyval.evaluate().unwrap());

View File

@@ -23,7 +23,7 @@ use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
use common_query::prelude::*;
use datatypes::prelude::*;
use datatypes::value::{ListValue, OrderedFloat};
use datatypes::vectors::{ConstantVector, Float64Vector, ListVector};
use datatypes::vectors::{ConstantVector, Float64Vector, Helper, ListVector};
use datatypes::with_match_primitive_type_id;
use num_traits::AsPrimitive;
use snafu::{ensure, OptionExt, ResultExt};
@@ -33,18 +33,12 @@ use statrs::statistics::Statistics;
// https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.norm.html
#[derive(Debug, Default)]
pub struct ScipyStatsNormCdf<T>
where
T: Primitive + AsPrimitive<f64> + std::iter::Sum<T>,
{
pub struct ScipyStatsNormCdf<T> {
values: Vec<T>,
x: Option<f64>,
}
impl<T> ScipyStatsNormCdf<T>
where
T: Primitive + AsPrimitive<f64> + std::iter::Sum<T>,
{
impl<T> ScipyStatsNormCdf<T> {
fn push(&mut self, value: T) {
self.values.push(value);
}
@@ -52,8 +46,8 @@ where
impl<T> Accumulator for ScipyStatsNormCdf<T>
where
T: Primitive + AsPrimitive<f64> + std::iter::Sum<T>,
for<'a> T: Scalar<RefType<'a> = T>,
T: WrapperType + std::iter::Sum<T>,
T::Native: AsPrimitive<f64>,
{
fn state(&self) -> Result<Vec<Value>> {
let nums = self
@@ -64,7 +58,7 @@ where
Ok(vec![
Value::List(ListValue::new(
Some(Box::new(nums)),
T::default().into().data_type(),
T::LogicalType::build_data_type(),
)),
self.x.into(),
])
@@ -86,14 +80,14 @@ where
let mut len = 1;
let column: &<T as Scalar>::VectorType = if column.is_const() {
len = column.len();
let column: &ConstantVector = unsafe { VectorHelper::static_cast(column) };
unsafe { VectorHelper::static_cast(column.inner()) }
let column: &ConstantVector = unsafe { Helper::static_cast(column) };
unsafe { Helper::static_cast(column.inner()) }
} else {
unsafe { VectorHelper::static_cast(column) }
unsafe { Helper::static_cast(column) }
};
let x = &values[1];
let x = VectorHelper::check_get_scalar::<f64>(x).context(error::InvalidInputsSnafu {
let x = Helper::check_get_scalar::<f64>(x).context(error::InvalidInputTypeSnafu {
err_msg: "expecting \"SCIPYSTATSNORMCDF\" function's second argument to be a positive integer",
})?;
let first = x.get(0);
@@ -160,19 +154,19 @@ where
),
})?;
for value in values.values_iter() {
let value = value.context(FromScalarValueSnafu)?;
let column: &<T as Scalar>::VectorType = unsafe { VectorHelper::static_cast(&value) };
for v in column.iter_data().flatten() {
self.push(v);
if let Some(value) = value.context(FromScalarValueSnafu)? {
let column: &<T as Scalar>::VectorType = unsafe { Helper::static_cast(&value) };
for v in column.iter_data().flatten() {
self.push(v);
}
}
}
Ok(())
}
fn evaluate(&self) -> Result<Value> {
let values = self.values.iter().map(|&v| v.as_()).collect::<Vec<_>>();
let mean = values.clone().mean();
let std_dev = values.std_dev();
let mean = self.values.iter().map(|v| v.into_native().as_()).mean();
let std_dev = self.values.iter().map(|v| v.into_native().as_()).std_dev();
if mean.is_nan() || std_dev.is_nan() {
Ok(Value::Null)
} else {
@@ -198,7 +192,7 @@ impl AggregateFunctionCreator for ScipyStatsNormCdfAccumulatorCreator {
with_match_primitive_type_id!(
input_type.logical_type_id(),
|$S| {
Ok(Box::new(ScipyStatsNormCdf::<$S>::default()))
Ok(Box::new(ScipyStatsNormCdf::<<$S as LogicalPrimitiveType>::Wrapper>::default()))
},
{
let err_msg = format!(
@@ -230,7 +224,7 @@ impl AggregateFunctionCreator for ScipyStatsNormCdfAccumulatorCreator {
#[cfg(test)]
mod test {
use datatypes::vectors::PrimitiveVector;
use datatypes::vectors::{Float64Vector, Int32Vector};
use super::*;
#[test]
@@ -244,12 +238,8 @@ mod test {
// test update no null-value batch
let mut scipy_stats_norm_cdf = ScipyStatsNormCdf::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![
Some(-1i32),
Some(1),
Some(2),
])),
Arc::new(PrimitiveVector::<f64>::from(vec![
Arc::new(Int32Vector::from(vec![Some(-1i32), Some(1), Some(2)])),
Arc::new(Float64Vector::from(vec![
Some(2.0_f64),
Some(2.0_f64),
Some(2.0_f64),
@@ -264,13 +254,8 @@ mod test {
// test update null-value batch
let mut scipy_stats_norm_cdf = ScipyStatsNormCdf::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![
Some(-2i32),
None,
Some(3),
Some(4),
])),
Arc::new(PrimitiveVector::<f64>::from(vec![
Arc::new(Int32Vector::from(vec![Some(-2i32), None, Some(3), Some(4)])),
Arc::new(Float64Vector::from(vec![
Some(2.0_f64),
None,
Some(2.0_f64),

View File

@@ -23,7 +23,7 @@ use common_query::logical_plan::{Accumulator, AggregateFunctionCreator};
use common_query::prelude::*;
use datatypes::prelude::*;
use datatypes::value::{ListValue, OrderedFloat};
use datatypes::vectors::{ConstantVector, Float64Vector, ListVector};
use datatypes::vectors::{ConstantVector, Float64Vector, Helper, ListVector};
use datatypes::with_match_primitive_type_id;
use num_traits::AsPrimitive;
use snafu::{ensure, OptionExt, ResultExt};
@@ -33,18 +33,12 @@ use statrs::statistics::Statistics;
// https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.norm.html
#[derive(Debug, Default)]
pub struct ScipyStatsNormPdf<T>
where
T: Primitive + AsPrimitive<f64> + std::iter::Sum<T>,
{
pub struct ScipyStatsNormPdf<T> {
values: Vec<T>,
x: Option<f64>,
}
impl<T> ScipyStatsNormPdf<T>
where
T: Primitive + AsPrimitive<f64> + std::iter::Sum<T>,
{
impl<T> ScipyStatsNormPdf<T> {
fn push(&mut self, value: T) {
self.values.push(value);
}
@@ -52,8 +46,8 @@ where
impl<T> Accumulator for ScipyStatsNormPdf<T>
where
T: Primitive + AsPrimitive<f64> + std::iter::Sum<T>,
for<'a> T: Scalar<RefType<'a> = T>,
T: WrapperType,
T::Native: AsPrimitive<f64> + std::iter::Sum<T>,
{
fn state(&self) -> Result<Vec<Value>> {
let nums = self
@@ -64,7 +58,7 @@ where
Ok(vec![
Value::List(ListValue::new(
Some(Box::new(nums)),
T::default().into().data_type(),
T::LogicalType::build_data_type(),
)),
self.x.into(),
])
@@ -86,14 +80,14 @@ where
let mut len = 1;
let column: &<T as Scalar>::VectorType = if column.is_const() {
len = column.len();
let column: &ConstantVector = unsafe { VectorHelper::static_cast(column) };
unsafe { VectorHelper::static_cast(column.inner()) }
let column: &ConstantVector = unsafe { Helper::static_cast(column) };
unsafe { Helper::static_cast(column.inner()) }
} else {
unsafe { VectorHelper::static_cast(column) }
unsafe { Helper::static_cast(column) }
};
let x = &values[1];
let x = VectorHelper::check_get_scalar::<f64>(x).context(error::InvalidInputsSnafu {
let x = Helper::check_get_scalar::<f64>(x).context(error::InvalidInputTypeSnafu {
err_msg: "expecting \"SCIPYSTATSNORMPDF\" function's second argument to be a positive integer",
})?;
let first = x.get(0);
@@ -160,19 +154,20 @@ where
),
})?;
for value in values.values_iter() {
let value = value.context(FromScalarValueSnafu)?;
let column: &<T as Scalar>::VectorType = unsafe { VectorHelper::static_cast(&value) };
for v in column.iter_data().flatten() {
self.push(v);
if let Some(value) = value.context(FromScalarValueSnafu)? {
let column: &<T as Scalar>::VectorType = unsafe { Helper::static_cast(&value) };
for v in column.iter_data().flatten() {
self.push(v);
}
}
}
Ok(())
}
fn evaluate(&self) -> Result<Value> {
let values = self.values.iter().map(|&v| v.as_()).collect::<Vec<_>>();
let mean = values.clone().mean();
let std_dev = values.std_dev();
let mean = self.values.iter().map(|v| v.into_native().as_()).mean();
let std_dev = self.values.iter().map(|v| v.into_native().as_()).std_dev();
if mean.is_nan() || std_dev.is_nan() {
Ok(Value::Null)
} else {
@@ -198,7 +193,7 @@ impl AggregateFunctionCreator for ScipyStatsNormPdfAccumulatorCreator {
with_match_primitive_type_id!(
input_type.logical_type_id(),
|$S| {
Ok(Box::new(ScipyStatsNormPdf::<$S>::default()))
Ok(Box::new(ScipyStatsNormPdf::<<$S as LogicalPrimitiveType>::Wrapper>::default()))
},
{
let err_msg = format!(
@@ -230,7 +225,7 @@ impl AggregateFunctionCreator for ScipyStatsNormPdfAccumulatorCreator {
#[cfg(test)]
mod test {
use datatypes::vectors::PrimitiveVector;
use datatypes::vectors::{Float64Vector, Int32Vector};
use super::*;
#[test]
@@ -244,12 +239,8 @@ mod test {
// test update no null-value batch
let mut scipy_stats_norm_pdf = ScipyStatsNormPdf::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![
Some(-1i32),
Some(1),
Some(2),
])),
Arc::new(PrimitiveVector::<f64>::from(vec![
Arc::new(Int32Vector::from(vec![Some(-1i32), Some(1), Some(2)])),
Arc::new(Float64Vector::from(vec![
Some(2.0_f64),
Some(2.0_f64),
Some(2.0_f64),
@@ -264,13 +255,8 @@ mod test {
// test update null-value batch
let mut scipy_stats_norm_pdf = ScipyStatsNormPdf::<i32>::default();
let v: Vec<VectorRef> = vec![
Arc::new(PrimitiveVector::<i32>::from(vec![
Some(-2i32),
None,
Some(3),
Some(4),
])),
Arc::new(PrimitiveVector::<f64>::from(vec![
Arc::new(Int32Vector::from(vec![Some(-2i32), None, Some(3), Some(4)])),
Arc::new(Float64Vector::from(vec![
Some(2.0_f64),
None,
Some(2.0_f64),

View File

@@ -14,10 +14,10 @@
use std::iter;
use common_query::error::Result;
use datatypes::prelude::*;
use datatypes::vectors::ConstantVector;
use datatypes::vectors::{ConstantVector, Helper};
use crate::error::Result;
use crate::scalars::expression::ctx::EvalContext;
pub fn scalar_binary_op<L: Scalar, R: Scalar, O: Scalar, F>(
@@ -36,10 +36,9 @@ where
let result = match (l.is_const(), r.is_const()) {
(false, true) => {
let left: &<L as Scalar>::VectorType = unsafe { VectorHelper::static_cast(l) };
let right: &ConstantVector = unsafe { VectorHelper::static_cast(r) };
let right: &<R as Scalar>::VectorType =
unsafe { VectorHelper::static_cast(right.inner()) };
let left: &<L as Scalar>::VectorType = unsafe { Helper::static_cast(l) };
let right: &ConstantVector = unsafe { Helper::static_cast(r) };
let right: &<R as Scalar>::VectorType = unsafe { Helper::static_cast(right.inner()) };
let b = right.get_data(0);
let it = left.iter_data().map(|a| f(a, b, ctx));
@@ -47,8 +46,8 @@ where
}
(false, false) => {
let left: &<L as Scalar>::VectorType = unsafe { VectorHelper::static_cast(l) };
let right: &<R as Scalar>::VectorType = unsafe { VectorHelper::static_cast(r) };
let left: &<L as Scalar>::VectorType = unsafe { Helper::static_cast(l) };
let right: &<R as Scalar>::VectorType = unsafe { Helper::static_cast(r) };
let it = left
.iter_data()
@@ -58,25 +57,22 @@ where
}
(true, false) => {
let left: &ConstantVector = unsafe { VectorHelper::static_cast(l) };
let left: &<L as Scalar>::VectorType =
unsafe { VectorHelper::static_cast(left.inner()) };
let left: &ConstantVector = unsafe { Helper::static_cast(l) };
let left: &<L as Scalar>::VectorType = unsafe { Helper::static_cast(left.inner()) };
let a = left.get_data(0);
let right: &<R as Scalar>::VectorType = unsafe { VectorHelper::static_cast(r) };
let right: &<R as Scalar>::VectorType = unsafe { Helper::static_cast(r) };
let it = right.iter_data().map(|b| f(a, b, ctx));
<O as Scalar>::VectorType::from_owned_iterator(it)
}
(true, true) => {
let left: &ConstantVector = unsafe { VectorHelper::static_cast(l) };
let left: &<L as Scalar>::VectorType =
unsafe { VectorHelper::static_cast(left.inner()) };
let left: &ConstantVector = unsafe { Helper::static_cast(l) };
let left: &<L as Scalar>::VectorType = unsafe { Helper::static_cast(left.inner()) };
let a = left.get_data(0);
let right: &ConstantVector = unsafe { VectorHelper::static_cast(r) };
let right: &<R as Scalar>::VectorType =
unsafe { VectorHelper::static_cast(right.inner()) };
let right: &ConstantVector = unsafe { Helper::static_cast(r) };
let right: &<R as Scalar>::VectorType = unsafe { Helper::static_cast(right.inner()) };
let b = right.get_data(0);
let it = iter::repeat(a)

View File

@@ -13,8 +13,7 @@
// limitations under the License.
use chrono_tz::Tz;
use crate::error::Error;
use common_query::error::Error;
pub struct EvalContext {
_tz: Tz,

View File

@@ -12,10 +12,11 @@
// See the License for the specific language governing permissions and
// limitations under the License.
use common_query::error::{self, Result};
use datatypes::prelude::*;
use datatypes::vectors::Helper;
use snafu::ResultExt;
use crate::error::{GetScalarVectorSnafu, Result};
use crate::scalars::expression::ctx::EvalContext;
/// TODO: remove the allow_unused when it's used.
@@ -28,7 +29,7 @@ pub fn scalar_unary_op<L: Scalar, O: Scalar, F>(
where
F: Fn(Option<L::RefType<'_>>, &mut EvalContext) -> Option<O>,
{
let left = VectorHelper::check_get_scalar::<L>(l).context(GetScalarVectorSnafu)?;
let left = Helper::check_get_scalar::<L>(l).context(error::GetScalarVectorSnafu)?;
let it = left.iter_data().map(|a| f(a, ctx));
let result = <O as Scalar>::VectorType::from_owned_iterator(it);

View File

@@ -16,12 +16,11 @@ use std::fmt;
use std::sync::Arc;
use chrono_tz::Tz;
use common_query::error::Result;
use common_query::prelude::Signature;
use datatypes::data_type::ConcreteDataType;
use datatypes::vectors::VectorRef;
use crate::error::Result;
#[derive(Clone)]
pub struct FunctionContext {
pub tz: Tz,

View File

@@ -13,10 +13,12 @@
// limitations under the License.
mod pow;
mod rate;
use std::sync::Arc;
pub use pow::PowFunction;
pub use rate::RateFunction;
use crate::scalars::function_registry::FunctionRegistry;
@@ -25,5 +27,6 @@ pub(crate) struct MathFunction;
impl MathFunction {
pub fn register(registry: &FunctionRegistry) {
registry.register(Arc::new(PowFunction::default()));
registry.register(Arc::new(RateFunction::default()))
}
}

View File

@@ -15,15 +15,16 @@
use std::fmt;
use std::sync::Arc;
use common_query::error::Result;
use common_query::prelude::{Signature, Volatility};
use datatypes::data_type::DataType;
use datatypes::prelude::ConcreteDataType;
use datatypes::types::LogicalPrimitiveType;
use datatypes::vectors::VectorRef;
use datatypes::with_match_primitive_type_id;
use num::traits::Pow;
use num_traits::AsPrimitive;
use crate::error::Result;
use crate::scalars::expression::{scalar_binary_op, EvalContext};
use crate::scalars::function::{Function, FunctionContext};
@@ -46,7 +47,7 @@ impl Function for PowFunction {
fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
with_match_primitive_type_id!(columns[0].data_type().logical_type_id(), |$S| {
with_match_primitive_type_id!(columns[1].data_type().logical_type_id(), |$T| {
let col = scalar_binary_op::<$S, $T, f64, _>(&columns[0], &columns[1], scalar_pow, &mut EvalContext::default())?;
let col = scalar_binary_op::<<$S as LogicalPrimitiveType>::Native, <$T as LogicalPrimitiveType>::Native, f64, _>(&columns[0], &columns[1], scalar_pow, &mut EvalContext::default())?;
Ok(Arc::new(col))
},{
unreachable!()

View File

@@ -0,0 +1,106 @@
// Copyright 2022 Greptime Team
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
use std::fmt;
use common_query::error::{self, Result};
use common_query::prelude::{Signature, Volatility};
use datatypes::arrow::compute::kernels::{arithmetic, cast};
use datatypes::arrow::datatypes::DataType;
use datatypes::prelude::*;
use datatypes::vectors::{Helper, VectorRef};
use snafu::ResultExt;
use crate::scalars::function::{Function, FunctionContext};
/// generates rates from a sequence of adjacent data points.
#[derive(Clone, Debug, Default)]
pub struct RateFunction;
impl fmt::Display for RateFunction {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "RATE")
}
}
impl Function for RateFunction {
fn name(&self) -> &str {
"prom_rate"
}
fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
Ok(ConcreteDataType::float64_datatype())
}
fn signature(&self) -> Signature {
Signature::uniform(2, ConcreteDataType::numerics(), Volatility::Immutable)
}
fn eval(&self, _func_ctx: FunctionContext, columns: &[VectorRef]) -> Result<VectorRef> {
let val = &columns[0].to_arrow_array();
let val_0 = val.slice(0, val.len() - 1);
let val_1 = val.slice(1, val.len() - 1);
let dv = arithmetic::subtract_dyn(&val_1, &val_0).context(error::ArrowComputeSnafu)?;
let ts = &columns[1].to_arrow_array();
let ts_0 = ts.slice(0, ts.len() - 1);
let ts_1 = ts.slice(1, ts.len() - 1);
let dt = arithmetic::subtract_dyn(&ts_1, &ts_0).context(error::ArrowComputeSnafu)?;
let dv = cast::cast(&dv, &DataType::Float64).context(error::TypeCastSnafu {
typ: DataType::Float64,
})?;
let dt = cast::cast(&dt, &DataType::Float64).context(error::TypeCastSnafu {
typ: DataType::Float64,
})?;
let rate = arithmetic::divide_dyn(&dv, &dt).context(error::ArrowComputeSnafu)?;
let v = Helper::try_into_vector(&rate).context(error::FromArrowArraySnafu)?;
Ok(v)
}
}
#[cfg(test)]
mod tests {
use std::sync::Arc;
use common_query::prelude::TypeSignature;
use datatypes::vectors::{Float32Vector, Float64Vector, Int64Vector};
use super::*;
#[test]
fn test_rate_function() {
let rate = RateFunction::default();
assert_eq!("prom_rate", rate.name());
assert_eq!(
ConcreteDataType::float64_datatype(),
rate.return_type(&[]).unwrap()
);
assert!(matches!(rate.signature(),
Signature {
type_signature: TypeSignature::Uniform(2, valid_types),
volatility: Volatility::Immutable
} if valid_types == ConcreteDataType::numerics()
));
let values = vec![1.0, 3.0, 6.0];
let ts = vec![0, 1, 2];
let args: Vec<VectorRef> = vec![
Arc::new(Float32Vector::from_vec(values)),
Arc::new(Int64Vector::from_vec(ts)),
];
let vector = rate.eval(FunctionContext::default(), &args).unwrap();
let expect: VectorRef = Arc::new(Float64Vector::from_vec(vec![2.0, 3.0]));
assert_eq!(expect, vector);
}
}

View File

@@ -13,7 +13,6 @@
// limitations under the License.
mod clip;
#[allow(unused)]
mod interp;
use std::sync::Arc;

View File

@@ -15,14 +15,15 @@
use std::fmt;
use std::sync::Arc;
use common_query::error::Result;
use common_query::prelude::{Signature, Volatility};
use datatypes::data_type::{ConcreteDataType, DataType};
use datatypes::prelude::{Scalar, VectorRef};
use datatypes::with_match_primitive_type_id;
use num_traits::AsPrimitive;
use datatypes::arrow::compute;
use datatypes::arrow::datatypes::ArrowPrimitiveType;
use datatypes::data_type::ConcreteDataType;
use datatypes::prelude::*;
use datatypes::vectors::PrimitiveVector;
use paste::paste;
use crate::error::Result;
use crate::scalars::expression::{scalar_binary_op, EvalContext};
use crate::scalars::function::{Function, FunctionContext};
@@ -34,25 +35,32 @@ macro_rules! define_eval {
($O: ident) => {
paste! {
fn [<eval_ $O>](columns: &[VectorRef]) -> Result<VectorRef> {
with_match_primitive_type_id!(columns[0].data_type().logical_type_id(), |$S| {
with_match_primitive_type_id!(columns[1].data_type().logical_type_id(), |$T| {
with_match_primitive_type_id!(columns[2].data_type().logical_type_id(), |$R| {
// clip(a, min, max) is equals to min(max(a, min), max)
let col: VectorRef = Arc::new(scalar_binary_op::<$S, $T, $O, _>(&columns[0], &columns[1], scalar_max, &mut EvalContext::default())?);
let col = scalar_binary_op::<$O, $R, $O, _>(&col, &columns[2], scalar_min, &mut EvalContext::default())?;
Ok(Arc::new(col))
}, {
unreachable!()
})
}, {
unreachable!()
})
}, {
unreachable!()
})
fn cast_vector(input: &VectorRef) -> VectorRef {
Arc::new(PrimitiveVector::<<$O as WrapperType>::LogicalType>::try_from_arrow_array(
compute::cast(&input.to_arrow_array(), &<<<$O as WrapperType>::LogicalType as LogicalPrimitiveType>::ArrowPrimitive as ArrowPrimitiveType>::DATA_TYPE).unwrap()
).unwrap()) as _
}
let operator_1 = cast_vector(&columns[0]);
let operator_2 = cast_vector(&columns[1]);
let operator_3 = cast_vector(&columns[2]);
// clip(a, min, max) is equals to min(max(a, min), max)
let col: VectorRef = Arc::new(scalar_binary_op::<$O, $O, $O, _>(
&operator_1,
&operator_2,
scalar_max,
&mut EvalContext::default(),
)?);
let col = scalar_binary_op::<$O, $O, $O, _>(
&col,
&operator_3,
scalar_min,
&mut EvalContext::default(),
)?;
Ok(Arc::new(col))
}
}
}
};
}
define_eval!(i64);
@@ -108,27 +116,23 @@ pub fn max<T: PartialOrd>(input: T, max: T) -> T {
}
#[inline]
fn scalar_min<S, T, O>(left: Option<S>, right: Option<T>, _ctx: &mut EvalContext) -> Option<O>
fn scalar_min<O>(left: Option<O>, right: Option<O>, _ctx: &mut EvalContext) -> Option<O>
where
S: AsPrimitive<O>,
T: AsPrimitive<O>,
O: Scalar + Copy + PartialOrd,
{
match (left, right) {
(Some(left), Some(right)) => Some(min(left.as_(), right.as_())),
(Some(left), Some(right)) => Some(min(left, right)),
_ => None,
}
}
#[inline]
fn scalar_max<S, T, O>(left: Option<S>, right: Option<T>, _ctx: &mut EvalContext) -> Option<O>
fn scalar_max<O>(left: Option<O>, right: Option<O>, _ctx: &mut EvalContext) -> Option<O>
where
S: AsPrimitive<O>,
T: AsPrimitive<O>,
O: Scalar + Copy + PartialOrd,
{
match (left, right) {
(Some(left), Some(right)) => Some(max(left.as_(), right.as_())),
(Some(left), Some(right)) => Some(max(left, right)),
_ => None,
}
}
@@ -143,11 +147,15 @@ impl fmt::Display for ClipFunction {
mod tests {
use common_query::prelude::TypeSignature;
use datatypes::value::Value;
use datatypes::vectors::{ConstantVector, Float32Vector, Int32Vector, UInt32Vector};
use datatypes::vectors::{
ConstantVector, Float32Vector, Int16Vector, Int32Vector, Int8Vector, UInt16Vector,
UInt32Vector, UInt8Vector,
};
use super::*;
#[test]
fn test_clip_function() {
fn test_clip_signature() {
let clip = ClipFunction::default();
assert_eq!("clip", clip.name());
@@ -190,16 +198,21 @@ mod tests {
volatility: Volatility::Immutable
} if valid_types == ConcreteDataType::numerics()
));
}
#[test]
fn test_clip_fn_signed() {
let clip = ClipFunction::default();
// eval with signed integers
let args: Vec<VectorRef> = vec![
Arc::new(Int32Vector::from_values(0..10)),
Arc::new(ConstantVector::new(
Arc::new(Int32Vector::from_vec(vec![3])),
Arc::new(Int8Vector::from_vec(vec![3])),
10,
)),
Arc::new(ConstantVector::new(
Arc::new(Int32Vector::from_vec(vec![6])),
Arc::new(Int16Vector::from_vec(vec![6])),
10,
)),
];
@@ -217,16 +230,21 @@ mod tests {
assert!(matches!(vector.get(i), Value::Int64(v) if v == 6));
}
}
}
#[test]
fn test_clip_fn_unsigned() {
let clip = ClipFunction::default();
// eval with unsigned integers
let args: Vec<VectorRef> = vec![
Arc::new(UInt32Vector::from_values(0..10)),
Arc::new(UInt8Vector::from_values(0..10)),
Arc::new(ConstantVector::new(
Arc::new(UInt32Vector::from_vec(vec![3])),
10,
)),
Arc::new(ConstantVector::new(
Arc::new(UInt32Vector::from_vec(vec![6])),
Arc::new(UInt16Vector::from_vec(vec![6])),
10,
)),
];
@@ -244,12 +262,17 @@ mod tests {
assert!(matches!(vector.get(i), Value::UInt64(v) if v == 6));
}
}
}
#[test]
fn test_clip_fn_float() {
let clip = ClipFunction::default();
// eval with floats
let args: Vec<VectorRef> = vec![
Arc::new(Int32Vector::from_values(0..10)),
Arc::new(Int8Vector::from_values(0..10)),
Arc::new(ConstantVector::new(
Arc::new(Int32Vector::from_vec(vec![3])),
Arc::new(UInt32Vector::from_vec(vec![3])),
10,
)),
Arc::new(ConstantVector::new(

View File

@@ -14,41 +14,18 @@
use std::sync::Arc;
use arrow::array::PrimitiveArray;
use arrow::compute::cast::primitive_to_primitive;
use arrow::datatypes::DataType::Float64;
use common_query::error::{self, Result};
use datatypes::arrow::compute::cast;
use datatypes::arrow::datatypes::DataType as ArrowDataType;
use datatypes::data_type::DataType;
use datatypes::prelude::ScalarVector;
use datatypes::type_id::LogicalTypeId;
use datatypes::value::Value;
use datatypes::vectors::{Float64Vector, PrimitiveVector, Vector, VectorRef};
use datatypes::{arrow, with_match_primitive_type_id};
use snafu::{ensure, Snafu};
#[derive(Debug, Snafu)]
pub enum Error {
#[snafu(display(
"The length of the args is not enough, expect at least: {}, have: {}",
expect,
actual,
))]
ArgsLenNotEnough { expect: usize, actual: usize },
#[snafu(display("The sample {} is empty", name))]
SampleEmpty { name: String },
#[snafu(display(
"The length of the len1: {} don't match the length of the len2: {}",
len1,
len2,
))]
LenNotEquals { len1: usize, len2: usize },
}
pub type Result<T> = std::result::Result<T, Error>;
use datatypes::vectors::{Float64Vector, Vector, VectorRef};
use datatypes::with_match_primitive_type_id;
use snafu::{ensure, ResultExt};
/* search the biggest number that smaller than x in xp */
fn linear_search_ascending_vector(x: Value, xp: &PrimitiveVector<f64>) -> usize {
fn linear_search_ascending_vector(x: Value, xp: &Float64Vector) -> usize {
for i in 0..xp.len() {
if x < xp.get(i) {
return i - 1;
@@ -58,7 +35,7 @@ fn linear_search_ascending_vector(x: Value, xp: &PrimitiveVector<f64>) -> usize
}
/* search the biggest number that smaller than x in xp */
fn binary_search_ascending_vector(key: Value, xp: &PrimitiveVector<f64>) -> usize {
fn binary_search_ascending_vector(key: Value, xp: &Float64Vector) -> usize {
let mut left = 0;
let mut right = xp.len();
/* If len <= 4 use linear search. */
@@ -77,27 +54,33 @@ fn binary_search_ascending_vector(key: Value, xp: &PrimitiveVector<f64>) -> usiz
left - 1
}
fn concrete_type_to_primitive_vector(arg: &VectorRef) -> Result<PrimitiveVector<f64>> {
fn concrete_type_to_primitive_vector(arg: &VectorRef) -> Result<Float64Vector> {
with_match_primitive_type_id!(arg.data_type().logical_type_id(), |$S| {
let tmp = arg.to_arrow_array();
let from = tmp.as_any().downcast_ref::<PrimitiveArray<$S>>().expect("cast failed");
let array = primitive_to_primitive(from, &Float64);
Ok(PrimitiveVector::new(array))
let array = cast(&tmp, &ArrowDataType::Float64).context(error::TypeCastSnafu {
typ: ArrowDataType::Float64,
})?;
// Safety: array has been cast to Float64Array.
Ok(Float64Vector::try_from_arrow_array(array).unwrap())
},{
unreachable!()
})
}
/// https://github.com/numpy/numpy/blob/b101756ac02e390d605b2febcded30a1da50cc2c/numpy/core/src/multiarray/compiled_base.c#L491
#[allow(unused)]
pub fn interp(args: &[VectorRef]) -> Result<VectorRef> {
let mut left = None;
let mut right = None;
ensure!(
args.len() >= 3,
ArgsLenNotEnoughSnafu {
expect: 3_usize,
actual: args.len()
error::InvalidFuncArgsSnafu {
err_msg: format!(
"The length of the args is not enough, expect at least: {}, have: {}",
3,
args.len()
),
}
);
@@ -109,9 +92,12 @@ pub fn interp(args: &[VectorRef]) -> Result<VectorRef> {
if args.len() > 3 {
ensure!(
args.len() == 5,
ArgsLenNotEnoughSnafu {
expect: 5_usize,
actual: args.len()
error::InvalidFuncArgsSnafu {
err_msg: format!(
"The length of the args is not enough, expect at least: {}, have: {}",
5,
args.len()
),
}
);
@@ -123,14 +109,32 @@ pub fn interp(args: &[VectorRef]) -> Result<VectorRef> {
.get_data(0);
}
ensure!(x.len() != 0, SampleEmptySnafu { name: "x" });
ensure!(xp.len() != 0, SampleEmptySnafu { name: "xp" });
ensure!(fp.len() != 0, SampleEmptySnafu { name: "fp" });
ensure!(
x.len() != 0,
error::InvalidFuncArgsSnafu {
err_msg: "The sample x is empty",
}
);
ensure!(
xp.len() != 0,
error::InvalidFuncArgsSnafu {
err_msg: "The sample xp is empty",
}
);
ensure!(
fp.len() != 0,
error::InvalidFuncArgsSnafu {
err_msg: "The sample fp is empty",
}
);
ensure!(
xp.len() == fp.len(),
LenNotEqualsSnafu {
len1: xp.len(),
len2: fp.len(),
error::InvalidFuncArgsSnafu {
err_msg: format!(
"The length of the len1: {} don't match the length of the len2: {}",
xp.len(),
fp.len()
),
}
);
@@ -147,7 +151,7 @@ pub fn interp(args: &[VectorRef]) -> Result<VectorRef> {
let res;
if xp.len() == 1 {
res = x
let datas = x
.iter_data()
.map(|x| {
if Value::from(x) < xp.get(0) {
@@ -158,7 +162,8 @@ pub fn interp(args: &[VectorRef]) -> Result<VectorRef> {
fp.get_data(0)
}
})
.collect::<Float64Vector>();
.collect::<Vec<_>>();
res = Float64Vector::from(datas);
} else {
let mut j = 0;
/* only pre-calculate slopes if there are relatively few of them. */
@@ -185,7 +190,7 @@ pub fn interp(args: &[VectorRef]) -> Result<VectorRef> {
}
slopes = Some(slopes_tmp);
}
res = x
let datas = x
.iter_data()
.map(|x| match x {
Some(xi) => {
@@ -248,7 +253,8 @@ pub fn interp(args: &[VectorRef]) -> Result<VectorRef> {
}
_ => None,
})
.collect::<Float64Vector>();
.collect::<Vec<_>>();
res = Float64Vector::from(datas);
}
Ok(Arc::new(res) as _)
}
@@ -257,8 +263,7 @@ pub fn interp(args: &[VectorRef]) -> Result<VectorRef> {
mod tests {
use std::sync::Arc;
use datatypes::prelude::ScalarVectorBuilder;
use datatypes::vectors::{Int32Vector, Int64Vector, PrimitiveVectorBuilder};
use datatypes::vectors::{Int32Vector, Int64Vector};
use super::*;
#[test]
@@ -341,12 +346,8 @@ mod tests {
assert!(matches!(vector.get(0), Value::Float64(v) if v==x[0] as f64));
// x=None output:Null
let input = [None, Some(0.0), Some(0.3)];
let mut builder = PrimitiveVectorBuilder::with_capacity(input.len());
for v in input {
builder.push(v);
}
let x = builder.finish();
let input = vec![None, Some(0.0), Some(0.3)];
let x = Float64Vector::from(input);
let args: Vec<VectorRef> = vec![
Arc::new(x),
Arc::new(Int64Vector::from_vec(xp)),

View File

@@ -15,11 +15,11 @@
use std::fmt;
use std::sync::Arc;
use common_query::error::Result;
use common_query::prelude::{Signature, Volatility};
use datatypes::data_type::ConcreteDataType;
use datatypes::prelude::VectorRef;
use crate::error::Result;
use crate::scalars::expression::{scalar_binary_op, EvalContext};
use crate::scalars::function::{Function, FunctionContext};

View File

@@ -17,16 +17,17 @@
use std::fmt;
use std::sync::Arc;
use arrow::compute::arithmetics;
use arrow::datatypes::DataType as ArrowDatatype;
use arrow::scalar::PrimitiveScalar;
use common_query::error::{IntoVectorSnafu, UnsupportedInputDataTypeSnafu};
use common_query::error::{
ArrowComputeSnafu, IntoVectorSnafu, Result, TypeCastSnafu, UnsupportedInputDataTypeSnafu,
};
use common_query::prelude::{Signature, Volatility};
use datatypes::arrow::compute;
use datatypes::arrow::datatypes::{DataType as ArrowDatatype, Int64Type};
use datatypes::data_type::DataType;
use datatypes::prelude::ConcreteDataType;
use datatypes::vectors::{TimestampVector, VectorRef};
use datatypes::vectors::{TimestampMillisecondVector, VectorRef};
use snafu::ResultExt;
use crate::error::Result;
use crate::scalars::function::{Function, FunctionContext};
#[derive(Clone, Debug, Default)]
@@ -40,7 +41,7 @@ impl Function for FromUnixtimeFunction {
}
fn return_type(&self, _input_types: &[ConcreteDataType]) -> Result<ConcreteDataType> {
Ok(ConcreteDataType::timestamp_millis_datatype())
Ok(ConcreteDataType::timestamp_millisecond_datatype())
}
fn signature(&self) -> Signature {
@@ -56,14 +57,18 @@ impl Function for FromUnixtimeFunction {
ConcreteDataType::Int64(_) => {
let array = columns[0].to_arrow_array();
// Our timestamp vector's time unit is millisecond
let array = arithmetics::mul_scalar(
&*array,
&PrimitiveScalar::new(ArrowDatatype::Int64, Some(1000i64)),
);
let array = compute::multiply_scalar_dyn::<Int64Type>(&array, 1000i64)
.context(ArrowComputeSnafu)?;
let arrow_datatype = &self.return_type(&[]).unwrap().as_arrow_type();
Ok(Arc::new(
TimestampVector::try_from_arrow_array(array).context(IntoVectorSnafu {
data_type: ArrowDatatype::Int64,
TimestampMillisecondVector::try_from_arrow_array(
compute::cast(&array, arrow_datatype).context(TypeCastSnafu {
typ: ArrowDatatype::Int64,
})?,
)
.context(IntoVectorSnafu {
data_type: arrow_datatype.clone(),
})?,
))
}
@@ -71,8 +76,7 @@ impl Function for FromUnixtimeFunction {
function: NAME,
datatypes: columns.iter().map(|c| c.data_type()).collect::<Vec<_>>(),
}
.fail()
.map_err(|e| e.into()),
.fail(),
}
}
}
@@ -96,7 +100,7 @@ mod tests {
let f = FromUnixtimeFunction::default();
assert_eq!("from_unixtime", f.name());
assert_eq!(
ConcreteDataType::timestamp_millis_datatype(),
ConcreteDataType::timestamp_millisecond_datatype(),
f.return_type(&[]).unwrap()
);

View File

@@ -19,7 +19,8 @@ use common_query::prelude::{
ColumnarValue, ReturnTypeFunction, ScalarFunctionImplementation, ScalarUdf, ScalarValue,
};
use datatypes::error::Error as DataTypeError;
use datatypes::prelude::{ConcreteDataType, VectorHelper};
use datatypes::prelude::*;
use datatypes::vectors::Helper;
use snafu::ResultExt;
use crate::scalars::function::{FunctionContext, FunctionRef};
@@ -47,7 +48,7 @@ pub fn create_udf(func: FunctionRef) -> ScalarUdf {
let args: Result<Vec<_>, DataTypeError> = args
.iter()
.map(|arg| match arg {
ColumnarValue::Scalar(v) => VectorHelper::try_from_scalar_value(v.clone(), rows),
ColumnarValue::Scalar(v) => Helper::try_from_scalar_value(v.clone(), rows),
ColumnarValue::Vector(v) => Ok(v.clone()),
})
.collect();

View File

@@ -1,5 +1,5 @@
[package]
name = "common-insert"
name = "common-grpc-expr"
version = "0.1.0"
edition = "2021"
license = "Apache-2.0"
@@ -8,10 +8,12 @@ license = "Apache-2.0"
api = { path = "../../api" }
async-trait = "0.1"
common-base = { path = "../base" }
common-catalog = { path = "../catalog" }
common-error = { path = "../error" }
common-grpc = { path = "../grpc" }
common-query = { path = "../query" }
common-telemetry = { path = "../telemetry" }
common-time = { path = "../time" }
common-query = { path = "../query" }
datatypes = { path = "../../datatypes" }
snafu = { version = "0.7", features = ["backtraces"] }
table = { path = "../../table" }

View File

@@ -0,0 +1,234 @@
// Copyright 2022 Greptime Team
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
use std::sync::Arc;
use api::v1::alter_expr::Kind;
use api::v1::{AlterExpr, CreateExpr, DropColumns};
use common_catalog::consts::{DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME};
use datatypes::schema::{ColumnSchema, SchemaBuilder, SchemaRef};
use snafu::{ensure, OptionExt, ResultExt};
use table::metadata::TableId;
use table::requests::{AddColumnRequest, AlterKind, AlterTableRequest, CreateTableRequest};
use crate::error::{
ColumnNotFoundSnafu, CreateSchemaSnafu, InvalidColumnDefSnafu, MissingFieldSnafu,
MissingTimestampColumnSnafu, Result,
};
/// Convert an [`AlterExpr`] to an optional [`AlterTableRequest`]
pub fn alter_expr_to_request(expr: AlterExpr) -> Result<Option<AlterTableRequest>> {
match expr.kind {
Some(Kind::AddColumns(add_columns)) => {
let add_column_requests = add_columns
.add_columns
.into_iter()
.map(|ac| {
let column_def = ac.column_def.context(MissingFieldSnafu {
field: "column_def",
})?;
let schema =
column_def
.try_as_column_schema()
.context(InvalidColumnDefSnafu {
column: &column_def.name,
})?;
Ok(AddColumnRequest {
column_schema: schema,
is_key: ac.is_key,
})
})
.collect::<Result<Vec<_>>>()?;
let alter_kind = AlterKind::AddColumns {
columns: add_column_requests,
};
let request = AlterTableRequest {
catalog_name: expr.catalog_name,
schema_name: expr.schema_name,
table_name: expr.table_name,
alter_kind,
};
Ok(Some(request))
}
Some(Kind::DropColumns(DropColumns { drop_columns })) => {
let alter_kind = AlterKind::DropColumns {
names: drop_columns.into_iter().map(|c| c.name).collect(),
};
let request = AlterTableRequest {
catalog_name: expr.catalog_name,
schema_name: expr.schema_name,
table_name: expr.table_name,
alter_kind,
};
Ok(Some(request))
}
None => Ok(None),
}
}
pub fn create_table_schema(expr: &CreateExpr) -> Result<SchemaRef> {
let column_schemas = expr
.column_defs
.iter()
.map(|x| {
x.try_as_column_schema()
.context(InvalidColumnDefSnafu { column: &x.name })
})
.collect::<Result<Vec<ColumnSchema>>>()?;
ensure!(
column_schemas
.iter()
.any(|column| column.name == expr.time_index),
MissingTimestampColumnSnafu {
msg: format!("CreateExpr: {:?}", expr)
}
);
let column_schemas = column_schemas
.into_iter()
.map(|column_schema| {
if column_schema.name == expr.time_index {
column_schema.with_time_index(true)
} else {
column_schema
}
})
.collect::<Vec<_>>();
Ok(Arc::new(
SchemaBuilder::try_from(column_schemas)
.context(CreateSchemaSnafu)?
.build()
.context(CreateSchemaSnafu)?,
))
}
pub fn create_expr_to_request(table_id: TableId, expr: CreateExpr) -> Result<CreateTableRequest> {
let schema = create_table_schema(&expr)?;
let primary_key_indices = expr
.primary_keys
.iter()
.map(|key| {
schema
.column_index_by_name(key)
.context(ColumnNotFoundSnafu {
column_name: key,
table_name: &expr.table_name,
})
})
.collect::<Result<Vec<usize>>>()?;
let catalog_name = expr
.catalog_name
.unwrap_or_else(|| DEFAULT_CATALOG_NAME.to_string());
let schema_name = expr
.schema_name
.unwrap_or_else(|| DEFAULT_SCHEMA_NAME.to_string());
let region_ids = if expr.region_ids.is_empty() {
vec![0]
} else {
expr.region_ids
};
Ok(CreateTableRequest {
id: table_id,
catalog_name,
schema_name,
table_name: expr.table_name,
desc: expr.desc,
schema,
region_numbers: region_ids,
primary_key_indices,
create_if_not_exists: expr.create_if_not_exists,
table_options: expr.table_options,
})
}
#[cfg(test)]
mod tests {
use api::v1::{AddColumn, AddColumns, ColumnDataType, ColumnDef, DropColumn};
use datatypes::prelude::ConcreteDataType;
use super::*;
#[test]
fn test_alter_expr_to_request() {
let expr = AlterExpr {
catalog_name: None,
schema_name: None,
table_name: "monitor".to_string(),
kind: Some(Kind::AddColumns(AddColumns {
add_columns: vec![AddColumn {
column_def: Some(ColumnDef {
name: "mem_usage".to_string(),
datatype: ColumnDataType::Float64 as i32,
is_nullable: false,
default_constraint: None,
}),
is_key: false,
}],
})),
};
let alter_request = alter_expr_to_request(expr).unwrap().unwrap();
assert_eq!(None, alter_request.catalog_name);
assert_eq!(None, alter_request.schema_name);
assert_eq!("monitor".to_string(), alter_request.table_name);
let add_column = match alter_request.alter_kind {
AlterKind::AddColumns { mut columns } => columns.pop().unwrap(),
_ => unreachable!(),
};
assert!(!add_column.is_key);
assert_eq!("mem_usage", add_column.column_schema.name);
assert_eq!(
ConcreteDataType::float64_datatype(),
add_column.column_schema.data_type
);
}
#[test]
fn test_drop_column_expr() {
let expr = AlterExpr {
catalog_name: Some("test_catalog".to_string()),
schema_name: Some("test_schema".to_string()),
table_name: "monitor".to_string(),
kind: Some(Kind::DropColumns(DropColumns {
drop_columns: vec![DropColumn {
name: "mem_usage".to_string(),
}],
})),
};
let alter_request = alter_expr_to_request(expr).unwrap().unwrap();
assert_eq!(Some("test_catalog".to_string()), alter_request.catalog_name);
assert_eq!(Some("test_schema".to_string()), alter_request.schema_name);
assert_eq!("monitor".to_string(), alter_request.table_name);
let mut drop_names = match alter_request.alter_kind {
AlterKind::DropColumns { names } => names,
_ => unreachable!(),
};
assert_eq!(1, drop_names.len());
assert_eq!("mem_usage".to_string(), drop_names.pop().unwrap());
}
}

View File

@@ -22,7 +22,7 @@ use snafu::{Backtrace, ErrorCompat};
#[derive(Debug, Snafu)]
#[snafu(visibility(pub))]
pub enum Error {
#[snafu(display("Column {} not found in table {}", column_name, table_name))]
#[snafu(display("Column `{}` not found in table `{}`", column_name, table_name))]
ColumnNotFound {
column_name: String,
table_name: String,
@@ -57,8 +57,8 @@ pub enum Error {
backtrace: Backtrace,
},
#[snafu(display("Missing timestamp column in request"))]
MissingTimestampColumn { backtrace: Backtrace },
#[snafu(display("Missing timestamp column, msg: {}", msg))]
MissingTimestampColumn { msg: String, backtrace: Backtrace },
#[snafu(display("Invalid column proto: {}", err_msg))]
InvalidColumnProto {
@@ -70,6 +70,26 @@ pub enum Error {
#[snafu(backtrace)]
source: datatypes::error::Error,
},
#[snafu(display("Missing required field in protobuf, field: {}", field))]
MissingField { field: String, backtrace: Backtrace },
#[snafu(display("Invalid column default constraint, source: {}", source))]
ColumnDefaultConstraint {
#[snafu(backtrace)]
source: datatypes::error::Error,
},
#[snafu(display(
"Invalid column proto definition, column: {}, source: {}",
column,
source
))]
InvalidColumnDef {
column: String,
#[snafu(backtrace)]
source: api::error::Error,
},
}
pub type Result<T> = std::result::Result<T, Error>;
@@ -87,6 +107,9 @@ impl ErrorExt for Error {
| Error::MissingTimestampColumn { .. } => StatusCode::InvalidArguments,
Error::InvalidColumnProto { .. } => StatusCode::InvalidArguments,
Error::CreateVector { .. } => StatusCode::InvalidArguments,
Error::MissingField { .. } => StatusCode::InvalidArguments,
Error::ColumnDefaultConstraint { source, .. } => source.status_code(),
Error::InvalidColumnDef { source, .. } => source.status_code(),
}
}
fn backtrace_opt(&self) -> Option<&Backtrace> {

View File

@@ -14,30 +14,27 @@
use std::collections::hash_map::Entry;
use std::collections::{HashMap, HashSet};
use std::ops::Deref;
use std::sync::Arc;
use api::helper::ColumnDataTypeWrapper;
use api::v1::codec::InsertBatch;
use api::v1::column::{SemanticType, Values};
use api::v1::{AddColumn, AddColumns, Column, ColumnDataType, ColumnDef, CreateExpr};
use common_base::BitVec;
use common_time::timestamp::Timestamp;
use common_time::{Date, DateTime};
use datatypes::data_type::ConcreteDataType;
use datatypes::data_type::{ConcreteDataType, DataType};
use datatypes::prelude::{ValueRef, VectorRef};
use datatypes::schema::SchemaRef;
use datatypes::value::Value;
use datatypes::vectors::VectorBuilder;
use datatypes::vectors::MutableVector;
use snafu::{ensure, OptionExt, ResultExt};
use table::metadata::TableId;
use table::requests::{AddColumnRequest, AlterKind, AlterTableRequest, InsertRequest};
use table::Table;
use crate::error::{
ColumnDataTypeSnafu, ColumnNotFoundSnafu, CreateVectorSnafu, DecodeInsertSnafu,
DuplicatedTimestampColumnSnafu, IllegalInsertDataSnafu, InvalidColumnProtoSnafu,
MissingTimestampColumnSnafu, Result,
ColumnDataTypeSnafu, ColumnNotFoundSnafu, CreateVectorSnafu, DuplicatedTimestampColumnSnafu,
IllegalInsertDataSnafu, InvalidColumnProtoSnafu, MissingTimestampColumnSnafu, Result,
};
const TAG_SEMANTIC_TYPE: i32 = SemanticType::Tag as i32;
const TIMESTAMP_SEMANTIC_TYPE: i32 = SemanticType::Timestamp as i32;
@@ -52,35 +49,25 @@ fn build_column_def(column_name: &str, datatype: i32, nullable: bool) -> ColumnD
}
}
pub fn find_new_columns(
schema: &SchemaRef,
insert_batches: &[InsertBatch],
) -> Result<Option<AddColumns>> {
pub fn find_new_columns(schema: &SchemaRef, columns: &[Column]) -> Result<Option<AddColumns>> {
let mut columns_to_add = Vec::default();
let mut new_columns: HashSet<String> = HashSet::default();
for InsertBatch { columns, row_count } in insert_batches {
if *row_count == 0 || columns.is_empty() {
continue;
}
for Column {
column_name,
semantic_type,
datatype,
..
} in columns
for Column {
column_name,
semantic_type,
datatype,
..
} in columns
{
if schema.column_schema_by_name(column_name).is_none() && !new_columns.contains(column_name)
{
if schema.column_schema_by_name(column_name).is_none()
&& !new_columns.contains(column_name)
{
let column_def = Some(build_column_def(column_name, *datatype, true));
columns_to_add.push(AddColumn {
column_def,
is_key: *semantic_type == TAG_SEMANTIC_TYPE,
});
new_columns.insert(column_name.to_string());
}
let column_def = Some(build_column_def(column_name, *datatype, true));
columns_to_add.push(AddColumn {
column_def,
is_key: *semantic_type == TAG_SEMANTIC_TYPE,
});
new_columns.insert(column_name.to_string());
}
}
@@ -112,7 +99,7 @@ pub fn column_to_vector(column: &Column, rows: u32) -> Result<VectorRef> {
let column_datatype = wrapper.datatype();
let rows = rows as usize;
let mut vector = VectorBuilder::with_capacity(wrapper.into(), rows);
let mut vector = ConcreteDataType::from(wrapper).create_mutable_vector(rows);
if let Some(values) = &column.values {
let values = collect_column_values(column_datatype, values);
@@ -123,21 +110,31 @@ pub fn column_to_vector(column: &Column, rows: u32) -> Result<VectorRef> {
for i in 0..rows {
if let Some(true) = nulls_iter.next() {
vector.push_null();
vector
.push_value_ref(ValueRef::Null)
.context(CreateVectorSnafu)?;
} else {
let value_ref = values_iter.next().context(InvalidColumnProtoSnafu {
err_msg: format!(
"value not found at position {} of column {}",
i, &column.column_name
),
})?;
vector.try_push_ref(value_ref).context(CreateVectorSnafu)?;
let value_ref = values_iter
.next()
.with_context(|| InvalidColumnProtoSnafu {
err_msg: format!(
"value not found at position {} of column {}",
i, &column.column_name
),
})?;
vector
.push_value_ref(value_ref)
.context(CreateVectorSnafu)?;
}
}
} else {
(0..rows).for_each(|_| vector.push_null());
(0..rows).try_for_each(|_| {
vector
.push_value_ref(ValueRef::Null)
.context(CreateVectorSnafu)
})?;
}
Ok(vector.finish())
Ok(vector.to_vector())
}
fn collect_column_values(column_datatype: ColumnDataType, values: &Values) -> Vec<ValueRef> {
@@ -187,9 +184,24 @@ fn collect_column_values(column_datatype: ColumnDataType, values: &Values) -> Ve
DateTime::new(*v)
))
}
ColumnDataType::Timestamp => {
collect_values!(values.ts_millis_values, |v| ValueRef::Timestamp(
Timestamp::from_millis(*v)
ColumnDataType::TimestampSecond => {
collect_values!(values.ts_second_values, |v| ValueRef::Timestamp(
Timestamp::new_second(*v)
))
}
ColumnDataType::TimestampMillisecond => {
collect_values!(values.ts_millisecond_values, |v| ValueRef::Timestamp(
Timestamp::new_millisecond(*v)
))
}
ColumnDataType::TimestampMicrosecond => {
collect_values!(values.ts_millisecond_values, |v| ValueRef::Timestamp(
Timestamp::new_microsecond(*v)
))
}
ColumnDataType::TimestampNanosecond => {
collect_values!(values.ts_millisecond_values, |v| ValueRef::Timestamp(
Timestamp::new_nanosecond(*v)
))
}
}
@@ -201,89 +213,84 @@ pub fn build_create_expr_from_insertion(
schema_name: &str,
table_id: Option<TableId>,
table_name: &str,
insert_batches: &[InsertBatch],
columns: &[Column],
) -> Result<CreateExpr> {
let mut new_columns: HashSet<String> = HashSet::default();
let mut column_defs = Vec::default();
let mut primary_key_indices = Vec::default();
let mut timestamp_index = usize::MAX;
for InsertBatch { columns, row_count } in insert_batches {
if *row_count == 0 || columns.is_empty() {
continue;
}
for Column {
column_name,
semantic_type,
datatype,
..
} in columns
{
if !new_columns.contains(column_name) {
let mut is_nullable = true;
match *semantic_type {
TAG_SEMANTIC_TYPE => primary_key_indices.push(column_defs.len()),
TIMESTAMP_SEMANTIC_TYPE => {
ensure!(
timestamp_index == usize::MAX,
DuplicatedTimestampColumnSnafu {
exists: &columns[timestamp_index].column_name,
duplicated: column_name,
}
);
timestamp_index = column_defs.len();
// Timestamp column must not be null.
is_nullable = false;
}
_ => {}
for Column {
column_name,
semantic_type,
datatype,
..
} in columns
{
if !new_columns.contains(column_name) {
let mut is_nullable = true;
match *semantic_type {
TAG_SEMANTIC_TYPE => primary_key_indices.push(column_defs.len()),
TIMESTAMP_SEMANTIC_TYPE => {
ensure!(
timestamp_index == usize::MAX,
DuplicatedTimestampColumnSnafu {
exists: &columns[timestamp_index].column_name,
duplicated: column_name,
}
);
timestamp_index = column_defs.len();
// Timestamp column must not be null.
is_nullable = false;
}
let column_def = build_column_def(column_name, *datatype, is_nullable);
column_defs.push(column_def);
new_columns.insert(column_name.to_string());
_ => {}
}
let column_def = build_column_def(column_name, *datatype, is_nullable);
column_defs.push(column_def);
new_columns.insert(column_name.to_string());
}
ensure!(timestamp_index != usize::MAX, MissingTimestampColumnSnafu);
let timestamp_field_name = columns[timestamp_index].column_name.clone();
let primary_keys = primary_key_indices
.iter()
.map(|idx| columns[*idx].column_name.clone())
.collect::<Vec<_>>();
let expr = CreateExpr {
catalog_name: Some(catalog_name.to_string()),
schema_name: Some(schema_name.to_string()),
table_name: table_name.to_string(),
desc: Some("Created on insertion".to_string()),
column_defs,
time_index: timestamp_field_name,
primary_keys,
create_if_not_exists: true,
table_options: Default::default(),
table_id,
region_ids: vec![0], // TODO:(hl): region id should be allocated by frontend
};
return Ok(expr);
}
IllegalInsertDataSnafu.fail()
ensure!(
timestamp_index != usize::MAX,
MissingTimestampColumnSnafu { msg: table_name }
);
let timestamp_field_name = columns[timestamp_index].column_name.clone();
let primary_keys = primary_key_indices
.iter()
.map(|idx| columns[*idx].column_name.clone())
.collect::<Vec<_>>();
let expr = CreateExpr {
catalog_name: Some(catalog_name.to_string()),
schema_name: Some(schema_name.to_string()),
table_name: table_name.to_string(),
desc: Some("Created on insertion".to_string()),
column_defs,
time_index: timestamp_field_name,
primary_keys,
create_if_not_exists: true,
table_options: Default::default(),
table_id,
region_ids: vec![0], // TODO:(hl): region id should be allocated by frontend
};
Ok(expr)
}
pub fn insertion_expr_to_request(
catalog_name: &str,
schema_name: &str,
table_name: &str,
insert_batches: Vec<InsertBatch>,
insert_batches: Vec<(Vec<Column>, u32)>,
table: Arc<dyn Table>,
) -> Result<InsertRequest> {
let schema = table.schema();
let mut columns_builders = HashMap::with_capacity(schema.column_schemas().len());
for InsertBatch { columns, row_count } in insert_batches {
for (columns, row_count) in insert_batches {
for Column {
column_name,
values,
@@ -307,10 +314,7 @@ pub fn insertion_expr_to_request(
},
)?;
let data_type = &column_schema.data_type;
entry.insert(VectorBuilder::with_capacity(
data_type.clone(),
row_count as usize,
))
entry.insert(data_type.create_mutable_vector(row_count as usize))
}
};
add_values_to_builder(vector_builder, values, row_count as usize, null_mask)?;
@@ -318,7 +322,7 @@ pub fn insertion_expr_to_request(
}
let columns_values = columns_builders
.into_iter()
.map(|(column_name, mut vector_builder)| (column_name, vector_builder.finish()))
.map(|(column_name, mut vector_builder)| (column_name, vector_builder.to_vector()))
.collect();
Ok(InsertRequest {
@@ -329,16 +333,8 @@ pub fn insertion_expr_to_request(
})
}
#[inline]
pub fn insert_batches(bytes_vec: &[Vec<u8>]) -> Result<Vec<InsertBatch>> {
bytes_vec
.iter()
.map(|bytes| bytes.deref().try_into().context(DecodeInsertSnafu))
.collect()
}
fn add_values_to_builder(
builder: &mut VectorBuilder,
builder: &mut Box<dyn MutableVector>,
values: Values,
row_count: usize,
null_mask: Vec<u8>,
@@ -349,9 +345,11 @@ fn add_values_to_builder(
if null_mask.is_empty() {
ensure!(values.len() == row_count, IllegalInsertDataSnafu);
values.iter().for_each(|value| {
builder.push(value);
});
values.iter().try_for_each(|value| {
builder
.push_value_ref(value.as_value_ref())
.context(CreateVectorSnafu)
})?;
} else {
let null_mask = BitVec::from_vec(null_mask);
ensure!(
@@ -362,9 +360,13 @@ fn add_values_to_builder(
let mut idx_of_values = 0;
for idx in 0..row_count {
match is_null(&null_mask, idx) {
Some(true) => builder.push(&Value::Null),
Some(true) => builder
.push_value_ref(ValueRef::Null)
.context(CreateVectorSnafu)?,
_ => {
builder.push(&values[idx_of_values]);
builder
.push_value_ref(values[idx_of_values].as_value_ref())
.context(CreateVectorSnafu)?;
idx_of_values += 1
}
}
@@ -444,9 +446,9 @@ fn convert_values(data_type: &ConcreteDataType, values: Values) -> Vec<Value> {
.map(|v| Value::Date(v.into()))
.collect(),
ConcreteDataType::Timestamp(_) => values
.ts_millis_values
.ts_millisecond_values
.into_iter()
.map(|v| Value::Timestamp(Timestamp::from_millis(v)))
.map(|v| Value::Timestamp(Timestamp::new_millisecond(v)))
.collect(),
ConcreteDataType::Null(_) => unreachable!(),
ConcreteDataType::List(_) => unreachable!(),
@@ -463,9 +465,8 @@ mod tests {
use std::sync::Arc;
use api::helper::ColumnDataTypeWrapper;
use api::v1::codec::InsertBatch;
use api::v1::column::{self, SemanticType, Values};
use api::v1::{insert_expr, Column, ColumnDataType};
use api::v1::{Column, ColumnDataType};
use common_base::BitVec;
use common_query::physical_plan::PhysicalPlanRef;
use common_query::prelude::Expr;
@@ -479,11 +480,12 @@ mod tests {
use table::Table;
use super::{
build_create_expr_from_insertion, convert_values, find_new_columns, insert_batches,
insertion_expr_to_request, is_null, TAG_SEMANTIC_TYPE, TIMESTAMP_SEMANTIC_TYPE,
build_create_expr_from_insertion, convert_values, insertion_expr_to_request, is_null,
TAG_SEMANTIC_TYPE, TIMESTAMP_SEMANTIC_TYPE,
};
use crate::error;
use crate::error::ColumnDataTypeSnafu;
use crate::insert::find_new_columns;
#[inline]
fn build_column_schema(
@@ -508,11 +510,10 @@ mod tests {
assert!(build_create_expr_from_insertion("", "", table_id, table_name, &[]).is_err());
let mock_batch_bytes = mock_insert_batches();
let insert_batches = insert_batches(&mock_batch_bytes).unwrap();
let insert_batch = mock_insert_batch();
let create_expr =
build_create_expr_from_insertion("", "", table_id, table_name, &insert_batches)
build_create_expr_from_insertion("", "", table_id, table_name, &insert_batch.0)
.unwrap();
assert_eq!(table_id, create_expr.table_id);
@@ -570,7 +571,7 @@ mod tests {
);
assert_eq!(
ConcreteDataType::timestamp_millis_datatype(),
ConcreteDataType::timestamp_millisecond_datatype(),
ConcreteDataType::from(
ColumnDataTypeWrapper::try_new(
column_defs
@@ -598,9 +599,9 @@ mod tests {
assert!(find_new_columns(&schema, &[]).unwrap().is_none());
let mock_insert_bytes = mock_insert_batches();
let insert_batches = insert_batches(&mock_insert_bytes).unwrap();
let add_columns = find_new_columns(&schema, &insert_batches).unwrap().unwrap();
let insert_batch = mock_insert_batch();
let add_columns = find_new_columns(&schema, &insert_batch.0).unwrap().unwrap();
assert_eq!(2, add_columns.add_columns.len());
let host_column = &add_columns.add_columns[0];
@@ -630,10 +631,7 @@ mod tests {
fn test_insertion_expr_to_request() {
let table: Arc<dyn Table> = Arc::new(DemoTable {});
let values = insert_expr::Values {
values: mock_insert_batches(),
};
let insert_batches = insert_batches(&values.values).unwrap();
let insert_batches = vec![mock_insert_batch()];
let insert_req =
insertion_expr_to_request("greptime", "public", "demo", insert_batches, table).unwrap();
@@ -654,8 +652,8 @@ mod tests {
assert_eq!(Value::Float64(0.1.into()), memory.get(1));
let ts = insert_req.columns_values.get("ts").unwrap();
assert_eq!(Value::Timestamp(Timestamp::from_millis(100)), ts.get(0));
assert_eq!(Value::Timestamp(Timestamp::from_millis(101)), ts.get(1));
assert_eq!(Value::Timestamp(Timestamp::new_millisecond(100)), ts.get(0));
assert_eq!(Value::Timestamp(Timestamp::new_millisecond(101)), ts.get(1));
}
#[test]
@@ -705,8 +703,12 @@ mod tests {
ColumnSchema::new("host", ConcreteDataType::string_datatype(), false),
ColumnSchema::new("cpu", ConcreteDataType::float64_datatype(), true),
ColumnSchema::new("memory", ConcreteDataType::float64_datatype(), true),
ColumnSchema::new("ts", ConcreteDataType::timestamp_millis_datatype(), true)
.with_time_index(true),
ColumnSchema::new(
"ts",
ConcreteDataType::timestamp_millisecond_datatype(),
true,
)
.with_time_index(true),
];
Arc::new(
@@ -731,7 +733,7 @@ mod tests {
}
}
fn mock_insert_batches() -> Vec<Vec<u8>> {
fn mock_insert_batch() -> (Vec<Column>, u32) {
let row_count = 2;
let host_vals = column::Values {
@@ -771,7 +773,7 @@ mod tests {
};
let ts_vals = column::Values {
ts_millis_values: vec![100, 101],
ts_millisecond_values: vec![100, 101],
..Default::default()
};
let ts_column = Column {
@@ -779,13 +781,12 @@ mod tests {
semantic_type: TIMESTAMP_SEMANTIC_TYPE,
values: Some(ts_vals),
null_mask: vec![0],
datatype: ColumnDataType::Timestamp as i32,
datatype: ColumnDataType::TimestampMillisecond as i32,
};
let insert_batch = InsertBatch {
columns: vec![host_column, cpu_column, mem_column, ts_column],
(
vec![host_column, cpu_column, mem_column, ts_column],
row_count,
};
vec![insert_batch.into()]
)
}
}

View File

@@ -1,3 +1,4 @@
#![feature(assert_matches)]
// Copyright 2022 Greptime Team
//
// Licensed under the Apache License, Version 2.0 (the "License");
@@ -12,9 +13,12 @@
// See the License for the specific language governing permissions and
// limitations under the License.
mod alter;
pub mod error;
mod insert;
pub use alter::{alter_expr_to_request, create_expr_to_request, create_table_schema};
pub use insert::{
build_alter_table_request, build_create_expr_from_insertion, column_to_vector,
find_new_columns, insert_batches, insertion_expr_to_request,
find_new_columns, insertion_expr_to_request,
};

Some files were not shown because too many files have changed in this diff Show More