rust/lancedb
1
0
Fork 0
mirror of https://github.com/lancedb/lancedb.git synced 2026-10-24 12:42:26 +00:00
Code Issues Packages Projects Releases Wiki Activity

refactor: replace a spec by unsetting it, not by setting over it

Browse Source 
A spec records how the generations already written were homed, and only its
maintained index set could move. Accepting that one repeat means a caller who
passes a different sharding — not knowing a spec is installed — gets a success
that changed nothing of what they asked for.

`set_lsm_write_spec` refuses while a spec is installed, and the refusal names
`unset_lsm_write_spec`. Unset drains the open writer first, so the rows it held
reach base before the new spec installs.

The default intent still reaches a live writer when an index is created, which
is the case that needed no spec change at all.
This commit is contained in:
XYZhan
2026-10-01 16:57:04 -04:00
parent 414b7f7aba
commit eb9459e3be
5 changed files with 66 additions and 74 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/python/tests/test_lsm_write_spec.py
+7 -7
View File
@@ -52,9 +52,9 @@ def test_set_lsm_write_spec_validates(tmp_path):
with pytest.raises(Exception, match="num_buckets"):
table.set_lsm_write_spec(LsmWriteSpec.bucket("id", 1025))
# Happy path then mutation rejected.
# Happy path, then a set over the installed spec is refused.
table.set_lsm_write_spec(LsmWriteSpec.bucket("id", 4))
with pytest.raises(Exception, match="mutation"):
with pytest.raises(Exception, match="already set"):
table.set_lsm_write_spec(LsmWriteSpec.bucket("id", 8))
@@ -176,18 +176,18 @@ def test_get_lsm_write_spec(tmp_path):
table.unset_lsm_write_spec()
assert table.get_lsm_write_spec() is None
# Identity round-trips (column recovered from the schema). Leaving the
# maintained set to be inferred picks up the index on the table, so the
# spec reads back naming it rather than as "infer".
# Identity round-trips (column recovered from the schema). An unnamed
# maintained set reads back as None: what is stored is the intent to
# maintain every index, not the set it resolves to now.
table.set_lsm_write_spec(LsmWriteSpec.identity("id"))
spec = table.get_lsm_write_spec()
assert spec.spec_type == "identity"
assert spec.column == "id"
assert spec.maintained_indexes == [idx_name]
assert spec.maintained_indexes is None
table.unset_lsm_write_spec()
# Unsharded round-trips (no routing column). Opting out is distinct from
# the inferred default.
# the unnamed default.
table.set_lsm_write_spec(LsmWriteSpec.unsharded().with_maintained_indexes([]))
spec = table.get_lsm_write_spec()
assert spec.spec_type == "unsharded"