mirror of
https://github.com/GreptimeTeam/greptimedb.git
synced 2026-10-20 02:42:29 +00:00
2224 lines
228 KiB
HTML
2224 lines
228 KiB
HTML
<!DOCTYPE html><html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width, initial-scale=1.0"><meta name="generator" content="rustdoc"><meta name="description" content="Source of the Rust file `src/flow/src/batching_mode/task.rs`."><title>task.rs - source</title><script>if(window.location.protocol!=="file:")document.head.insertAdjacentHTML("beforeend","SourceSerif4-Regular-6b053e98.ttf.woff2,FiraSans-Italic-81dc35de.woff2,FiraSans-Regular-0fe48ade.woff2,FiraSans-MediumItalic-ccf7e434.woff2,FiraSans-Medium-e1aa3f0a.woff2,SourceCodePro-Regular-8badfe75.ttf.woff2,SourceCodePro-Semibold-aa29a496.ttf.woff2".split(",").map(f=>`<link rel="preload" as="font" type="font/woff2"href="../../../static.files/${f}">`).join(""))</script><link rel="stylesheet" href="../../../static.files/normalize-9960930a.css"><link rel="stylesheet" href="../../../static.files/rustdoc-17e0aaed.css"><meta name="rustdoc-vars" data-root-path="../../../" data-static-root-path="../../../static.files/" data-current-crate="flow" data-themes="" data-resource-suffix="" data-rustdoc-version="1.96.1 (31fca3adb 2026-06-26)" data-channel="1.96.1" data-search-js="search-b5634cc7.js" data-stringdex-js="stringdex-2da4960a.js" data-settings-js="settings-170eb4bf.js" ><script src="../../../static.files/storage-41dd4d93.js"></script><script defer src="../../../static.files/src-script-813739b1.js"></script><script defer src="../../../src-files.js"></script><script defer src="../../../static.files/main-5013f961.js"></script><noscript><link rel="stylesheet" href="../../../static.files/noscript-f7c3ffd8.css"></noscript><link rel="alternate icon" type="image/png" href="../../../static.files/favicon-32x32-eab170b8.png"><link rel="icon" type="image/svg+xml" href="../../../static.files/favicon-044be391.svg"></head><body class="rustdoc src"><a class="skip-main-content" href="#main-content">Skip to main content</a><!--[if lte IE 11]><div class="warning">This old browser is unsupported and will most likely display funky things.</div><![endif]--><nav class="sidebar"><div class="src-sidebar-title"><h2>Files</h2></div></nav><div class="sidebar-resizer" title="Drag to resize sidebar"></div><main><section id="main-content" class="content" tabindex="-1"><div class="main-heading"><h1><div class="sub-heading">flow/batching_mode/</div>task.rs</h1><rustdoc-toolbar></rustdoc-toolbar></div><div class="example-wrap digits-4"><pre class="rust"><code><a href=#1 id=1 data-nosnippet>1</a><span class="comment">// Copyright 2023 Greptime Team
|
|
<a href=#2 id=2 data-nosnippet>2</a>//
|
|
<a href=#3 id=3 data-nosnippet>3</a>// Licensed under the Apache License, Version 2.0 (the "License");
|
|
<a href=#4 id=4 data-nosnippet>4</a>// you may not use this file except in compliance with the License.
|
|
<a href=#5 id=5 data-nosnippet>5</a>// You may obtain a copy of the License at
|
|
<a href=#6 id=6 data-nosnippet>6</a>//
|
|
<a href=#7 id=7 data-nosnippet>7</a>// http://www.apache.org/licenses/LICENSE-2.0
|
|
<a href=#8 id=8 data-nosnippet>8</a>//
|
|
<a href=#9 id=9 data-nosnippet>9</a>// Unless required by applicable law or agreed to in writing, software
|
|
<a href=#10 id=10 data-nosnippet>10</a>// distributed under the License is distributed on an "AS IS" BASIS,
|
|
<a href=#11 id=11 data-nosnippet>11</a>// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
<a href=#12 id=12 data-nosnippet>12</a>// See the License for the specific language governing permissions and
|
|
<a href=#13 id=13 data-nosnippet>13</a>// limitations under the License.
|
|
<a href=#14 id=14 data-nosnippet>14</a>
|
|
<a href=#15 id=15 data-nosnippet>15</a></span><span class="kw">use </span>std::collections::{BTreeMap, BTreeSet, HashMap, HashSet};
|
|
<a href=#16 id=16 data-nosnippet>16</a><span class="kw">use </span>std::sync::{Arc, RwLock};
|
|
<a href=#17 id=17 data-nosnippet>17</a><span class="kw">use </span>std::time::{Duration, SystemTime, UNIX_EPOCH};
|
|
<a href=#18 id=18 data-nosnippet>18</a>
|
|
<a href=#19 id=19 data-nosnippet>19</a><span class="kw">use </span>api::v1::{CreateTableExpr, TableName};
|
|
<a href=#20 id=20 data-nosnippet>20</a><span class="kw">use </span>catalog::CatalogManagerRef;
|
|
<a href=#21 id=21 data-nosnippet>21</a><span class="kw">use </span>catalog::kvbackend::KvBackendCatalogManager;
|
|
<a href=#22 id=22 data-nosnippet>22</a><span class="kw">use </span>client::OutputWithMetrics;
|
|
<a href=#23 id=23 data-nosnippet>23</a><span class="kw">use </span>common_error::ext::BoxedError;
|
|
<a href=#24 id=24 data-nosnippet>24</a><span class="kw">use </span>common_meta::key::schema_name::SchemaNameKey;
|
|
<a href=#25 id=25 data-nosnippet>25</a><span class="kw">use </span>common_query::OutputData;
|
|
<a href=#26 id=26 data-nosnippet>26</a><span class="kw">use </span>common_query::logical_plan::breakup_insert_plan;
|
|
<a href=#27 id=27 data-nosnippet>27</a><span class="kw">use </span>common_telemetry::tracing::warn;
|
|
<a href=#28 id=28 data-nosnippet>28</a><span class="kw">use </span>common_telemetry::{debug, info};
|
|
<a href=#29 id=29 data-nosnippet>29</a><span class="kw">use </span>common_time::{TimeToLive, Timestamp};
|
|
<a href=#30 id=30 data-nosnippet>30</a><span class="kw">use </span>datafusion::datasource::DefaultTableSource;
|
|
<a href=#31 id=31 data-nosnippet>31</a><span class="kw">use </span>datafusion::sql::unparser::expr_to_sql;
|
|
<a href=#32 id=32 data-nosnippet>32</a><span class="kw">use </span>datafusion_common::tree_node::{Transformed, TreeNode};
|
|
<a href=#33 id=33 data-nosnippet>33</a><span class="kw">use </span>datafusion_common::utils::quote_identifier;
|
|
<a href=#34 id=34 data-nosnippet>34</a><span class="kw">use </span>datafusion_common::{DFSchemaRef, ScalarValue, TableReference};
|
|
<a href=#35 id=35 data-nosnippet>35</a><span class="kw">use </span>datafusion_expr::{DmlStatement, LogicalPlan, Projection, WriteOp, col, lit};
|
|
<a href=#36 id=36 data-nosnippet>36</a><span class="kw">use </span>datatypes::schema::Schema;
|
|
<a href=#37 id=37 data-nosnippet>37</a><span class="kw">use </span>datatypes::vectors::Helper;
|
|
<a href=#38 id=38 data-nosnippet>38</a><span class="kw">use </span>futures::TryStreamExt;
|
|
<a href=#39 id=39 data-nosnippet>39</a><span class="kw">use </span>query::QueryEngineRef;
|
|
<a href=#40 id=40 data-nosnippet>40</a><span class="kw">use </span>query::options::{
|
|
<a href=#41 id=41 data-nosnippet>41</a> FLOW_INCREMENTAL_AFTER_SEQS, FLOW_INCREMENTAL_MODE, FLOW_INCREMENTAL_MODE_SEQUENCE_RANGE,
|
|
<a href=#42 id=42 data-nosnippet>42</a> FLOW_RETURN_REGION_SEQ,
|
|
<a href=#43 id=43 data-nosnippet>43</a>};
|
|
<a href=#44 id=44 data-nosnippet>44</a><span class="kw">use </span>query::query_engine::DefaultSerializer;
|
|
<a href=#45 id=45 data-nosnippet>45</a><span class="kw">use </span>session::context::QueryContextRef;
|
|
<a href=#46 id=46 data-nosnippet>46</a><span class="kw">use </span>snafu::{OptionExt, ResultExt};
|
|
<a href=#47 id=47 data-nosnippet>47</a><span class="kw">use </span>sql::parsers::utils::is_tql;
|
|
<a href=#48 id=48 data-nosnippet>48</a><span class="kw">use </span>store_api::mito_engine_options::MERGE_MODE_KEY;
|
|
<a href=#49 id=49 data-nosnippet>49</a><span class="kw">use </span>substrait::{DFLogicalSubstraitConvertor, SubstraitPlan};
|
|
<a href=#50 id=50 data-nosnippet>50</a><span class="kw">use </span>table::TableRef;
|
|
<a href=#51 id=51 data-nosnippet>51</a><span class="kw">use </span>table::table::adapter::DfTableProviderAdapter;
|
|
<a href=#52 id=52 data-nosnippet>52</a><span class="kw">use </span>tokio::sync::oneshot::error::TryRecvError;
|
|
<a href=#53 id=53 data-nosnippet>53</a><span class="kw">use </span>tokio::sync::{Mutex, OwnedMutexGuard, oneshot};
|
|
<a href=#54 id=54 data-nosnippet>54</a><span class="kw">use </span>tokio::time::Instant;
|
|
<a href=#55 id=55 data-nosnippet>55</a>
|
|
<a href=#56 id=56 data-nosnippet>56</a><span class="kw">use </span><span class="kw">crate</span>::batching_mode::BatchingModeOptions;
|
|
<a href=#57 id=57 data-nosnippet>57</a><span class="kw">use </span><span class="kw">crate</span>::batching_mode::checkpoint::checkpoint_mode_label;
|
|
<a href=#58 id=58 data-nosnippet>58</a><span class="kw">use </span><span class="kw">crate</span>::batching_mode::eval_schedule::{EvalSchedule, select_due_scheduled_times};
|
|
<a href=#59 id=59 data-nosnippet>59</a><span class="kw">use </span><span class="kw">crate</span>::batching_mode::frontend_client::{FrontendClient, PeerDesc};
|
|
<a href=#60 id=60 data-nosnippet>60</a><span class="kw">use </span><span class="kw">crate</span>::batching_mode::state::{
|
|
<a href=#61 id=61 data-nosnippet>61</a> CheckpointMode, DirtyTimeWindows, FilterExprInfo, TaskState, to_df_literal,
|
|
<a href=#62 id=62 data-nosnippet>62</a>};
|
|
<a href=#63 id=63 data-nosnippet>63</a><span class="kw">use </span><span class="kw">crate</span>::batching_mode::table_creator::{QueryType, create_table_with_expr};
|
|
<a href=#64 id=64 data-nosnippet>64</a><span class="kw">use </span><span class="kw">crate</span>::batching_mode::time_window::TimeWindowExpr;
|
|
<a href=#65 id=65 data-nosnippet>65</a><span class="kw">use </span><span class="kw">crate</span>::batching_mode::utils::{
|
|
<a href=#66 id=66 data-nosnippet>66</a> AddFilterRewriter, ColumnMatcherRewriter, analyze_incremental_aggregate_plan, df_plan_to_sql,
|
|
<a href=#67 id=67 data-nosnippet>67</a> gen_plan_with_matching_schema_and_values, get_table_info_df_schema, sql_to_df_plan,
|
|
<a href=#68 id=68 data-nosnippet>68</a>};
|
|
<a href=#69 id=69 data-nosnippet>69</a><span class="kw">use </span><span class="kw">crate</span>::df_optimizer::apply_df_optimizer;
|
|
<a href=#70 id=70 data-nosnippet>70</a><span class="kw">use </span><span class="kw">crate</span>::error::{
|
|
<a href=#71 id=71 data-nosnippet>71</a> DatafusionSnafu, ExternalSnafu, InvalidQuerySnafu, SubstraitEncodeLogicalPlanSnafu,
|
|
<a href=#72 id=72 data-nosnippet>72</a> UnexpectedSnafu,
|
|
<a href=#73 id=73 data-nosnippet>73</a>};
|
|
<a href=#74 id=74 data-nosnippet>74</a><span class="kw">use </span><span class="kw">crate</span>::metrics::{
|
|
<a href=#75 id=75 data-nosnippet>75</a> METRIC_FLOW_BATCHING_ENGINE_ERROR_CNT, METRIC_FLOW_BATCHING_ENGINE_QUERY_TIME,
|
|
<a href=#76 id=76 data-nosnippet>76</a> METRIC_FLOW_BATCHING_ENGINE_SLOW_QUERY, METRIC_FLOW_BATCHING_ENGINE_START_QUERY_CNT,
|
|
<a href=#77 id=77 data-nosnippet>77</a> METRIC_FLOW_ROWS,
|
|
<a href=#78 id=78 data-nosnippet>78</a>};
|
|
<a href=#79 id=79 data-nosnippet>79</a><span class="kw">use </span>crate::{Error, FlowId};
|
|
<a href=#80 id=80 data-nosnippet>80</a>
|
|
<a href=#81 id=81 data-nosnippet>81</a><span class="kw">mod </span>ckpt;
|
|
<a href=#82 id=82 data-nosnippet>82</a><span class="kw">mod </span>inc;
|
|
<a href=#83 id=83 data-nosnippet>83</a>
|
|
<a href=#84 id=84 data-nosnippet>84</a><span class="doccomment">/// Returns the current wall-clock Unix timestamp in seconds.
|
|
<a href=#85 id=85 data-nosnippet>85</a></span><span class="kw">fn </span>wall_clock_unix_secs() -> i64 {
|
|
<a href=#86 id=86 data-nosnippet>86</a> SystemTime::now()
|
|
<a href=#87 id=87 data-nosnippet>87</a> .duration_since(UNIX_EPOCH)
|
|
<a href=#88 id=88 data-nosnippet>88</a> .unwrap_or_default()
|
|
<a href=#89 id=89 data-nosnippet>89</a> .as_secs() <span class="kw">as </span>i64
|
|
<a href=#90 id=90 data-nosnippet>90</a>}
|
|
<a href=#91 id=91 data-nosnippet>91</a>
|
|
<a href=#92 id=92 data-nosnippet>92</a><span class="doccomment">/// Initial scheduler cursor for `start_scheduled_loop`: exactly one interval
|
|
<a href=#93 id=93 data-nosnippet>93</a>/// before `start_secs` so the first due scheduled time is `start_secs` itself.
|
|
<a href=#94 id=94 data-nosnippet>94</a>///
|
|
<a href=#95 id=95 data-nosnippet>95</a>/// Fallible: a `start_secs - interval_secs` difference that does not fit in
|
|
<a href=#96 id=96 data-nosnippet>96</a>/// `i64` is an explicit error instead of a saturated cursor that would make
|
|
<a href=#97 id=97 data-nosnippet>97</a>/// the first due scheduled time `start_secs + interval_secs` and silently skip
|
|
<a href=#98 id=98 data-nosnippet>98</a>/// the `start_secs` tick.
|
|
<a href=#99 id=99 data-nosnippet>99</a></span><span class="kw">fn </span>initial_schedule_cursor(start_secs: i64, interval_secs: i64) -> <span class="prelude-ty">Result</span><i64, Error> {
|
|
<a href=#100 id=100 data-nosnippet>100</a> <span class="kw">let </span>cursor = i128::from(start_secs) - i128::from(interval_secs);
|
|
<a href=#101 id=101 data-nosnippet>101</a> i64::try_from(cursor).map_err(|<span class="kw">_</span>| {
|
|
<a href=#102 id=102 data-nosnippet>102</a> UnexpectedSnafu {
|
|
<a href=#103 id=103 data-nosnippet>103</a> reason: <span class="macro">format!</span>(
|
|
<a href=#104 id=104 data-nosnippet>104</a> <span class="string">"Cannot compute the initial eval schedule cursor one interval before start {start_secs} (interval={interval_secs}): {cursor} does not fit in i64"
|
|
<a href=#105 id=105 data-nosnippet>105</a> </span>),
|
|
<a href=#106 id=106 data-nosnippet>106</a> }
|
|
<a href=#107 id=107 data-nosnippet>107</a> .build()
|
|
<a href=#108 id=108 data-nosnippet>108</a> })
|
|
<a href=#109 id=109 data-nosnippet>109</a>}
|
|
<a href=#110 id=110 data-nosnippet>110</a>
|
|
<a href=#111 id=111 data-nosnippet>111</a><span class="doccomment">/// Whole seconds to sleep until the next scheduled time `next`, measured from
|
|
<a href=#112 id=112 data-nosnippet>112</a>/// the current wall clock `wall_now_secs`.
|
|
<a href=#113 id=113 data-nosnippet>113</a>///
|
|
<a href=#114 id=114 data-nosnippet>114</a>/// Fallible: `next` must be strictly after `wall_now_secs` and the difference
|
|
<a href=#115 id=115 data-nosnippet>115</a>/// must fit in `u64`. In practice `i64::MAX - i64::MIN` is exactly `u64::MAX`,
|
|
<a href=#116 id=116 data-nosnippet>116</a>/// so the difference always fits once `next > wall_now_secs`; the explicit
|
|
<a href=#117 id=117 data-nosnippet>117</a>/// error keeps the scheduled loop panic-free and wrap-free regardless.
|
|
<a href=#118 id=118 data-nosnippet>118</a></span><span class="kw">fn </span>sleep_delta_secs(next: i64, wall_now_secs: i64) -> <span class="prelude-ty">Result</span><u64, Error> {
|
|
<a href=#119 id=119 data-nosnippet>119</a> <span class="kw">let </span>delta = i128::from(next) - i128::from(wall_now_secs);
|
|
<a href=#120 id=120 data-nosnippet>120</a> u64::try_from(delta).map_err(|<span class="kw">_</span>| {
|
|
<a href=#121 id=121 data-nosnippet>121</a> UnexpectedSnafu {
|
|
<a href=#122 id=122 data-nosnippet>122</a> reason: <span class="macro">format!</span>(
|
|
<a href=#123 id=123 data-nosnippet>123</a> <span class="string">"Cannot sleep until the next scheduled time {next}: the delta from wall clock {wall_now_secs} is {delta} seconds, which does not fit in u64"
|
|
<a href=#124 id=124 data-nosnippet>124</a> </span>),
|
|
<a href=#125 id=125 data-nosnippet>125</a> }
|
|
<a href=#126 id=126 data-nosnippet>126</a> .build()
|
|
<a href=#127 id=127 data-nosnippet>127</a> })
|
|
<a href=#128 id=128 data-nosnippet>128</a>}
|
|
<a href=#129 id=129 data-nosnippet>129</a>
|
|
<a href=#130 id=130 data-nosnippet>130</a><span class="doccomment">/// Scheduled time in seconds converted to milliseconds for the
|
|
<a href=#131 id=131 data-nosnippet>131</a>/// `FLOW_SCHEDULED_TIME_MILLIS` extension.
|
|
<a href=#132 id=132 data-nosnippet>132</a>///
|
|
<a href=#133 id=133 data-nosnippet>133</a>/// Fallible: a seconds value whose millisecond product does not fit in `i64`
|
|
<a href=#134 id=134 data-nosnippet>134</a>/// is an explicit error instead of a saturated `i64::MAX` that would silently
|
|
<a href=#135 id=135 data-nosnippet>135</a>/// misrepresent the logical scheduled time.
|
|
<a href=#136 id=136 data-nosnippet>136</a></span><span class="kw">fn </span>scheduled_time_millis(scheduled_time_secs: i64) -> <span class="prelude-ty">Result</span><i64, Error> {
|
|
<a href=#137 id=137 data-nosnippet>137</a> scheduled_time_secs.checked_mul(<span class="number">1000</span>).ok_or_else(|| {
|
|
<a href=#138 id=138 data-nosnippet>138</a> UnexpectedSnafu {
|
|
<a href=#139 id=139 data-nosnippet>139</a> reason: <span class="macro">format!</span>(
|
|
<a href=#140 id=140 data-nosnippet>140</a> <span class="string">"Cannot convert scheduled time {scheduled_time_secs}s to milliseconds: the product exceeds i64"
|
|
<a href=#141 id=141 data-nosnippet>141</a> </span>),
|
|
<a href=#142 id=142 data-nosnippet>142</a> }
|
|
<a href=#143 id=143 data-nosnippet>143</a> .build()
|
|
<a href=#144 id=144 data-nosnippet>144</a> })
|
|
<a href=#145 id=145 data-nosnippet>145</a>}
|
|
<a href=#146 id=146 data-nosnippet>146</a>
|
|
<a href=#147 id=147 data-nosnippet>147</a><span class="doccomment">/// The task's config, immutable once created
|
|
<a href=#148 id=148 data-nosnippet>148</a></span><span class="attr">#[derive(Clone)]
|
|
<a href=#149 id=149 data-nosnippet>149</a></span><span class="kw">pub struct </span>TaskConfig {
|
|
<a href=#150 id=150 data-nosnippet>150</a> <span class="kw">pub </span>flow_id: FlowId,
|
|
<a href=#151 id=151 data-nosnippet>151</a> <span class="kw">pub </span>query: String,
|
|
<a href=#152 id=152 data-nosnippet>152</a> <span class="doccomment">/// output schema of the query
|
|
<a href=#153 id=153 data-nosnippet>153</a> </span><span class="kw">pub </span>output_schema: DFSchemaRef,
|
|
<a href=#154 id=154 data-nosnippet>154</a> <span class="kw">pub </span>time_window_expr: <span class="prelude-ty">Option</span><TimeWindowExpr>,
|
|
<a href=#155 id=155 data-nosnippet>155</a> <span class="doccomment">/// in seconds
|
|
<a href=#156 id=156 data-nosnippet>156</a> </span><span class="kw">pub </span>expire_after: <span class="prelude-ty">Option</span><i64>,
|
|
<a href=#157 id=157 data-nosnippet>157</a> <span class="kw">pub </span>sink_table_name: [String; <span class="number">3</span>],
|
|
<a href=#158 id=158 data-nosnippet>158</a> <span class="kw">pub </span>source_table_names: HashSet<[String; <span class="number">3</span>]>,
|
|
<a href=#159 id=159 data-nosnippet>159</a> <span class="kw">pub </span>catalog_manager: CatalogManagerRef,
|
|
<a href=#160 id=160 data-nosnippet>160</a> <span class="kw">pub </span>query_type: QueryType,
|
|
<a href=#161 id=161 data-nosnippet>161</a> <span class="kw">pub </span>batch_opts: Arc<BatchingModeOptions>,
|
|
<a href=#162 id=162 data-nosnippet>162</a> <span class="kw">pub </span>exact_sequence_range_required: bool,
|
|
<a href=#163 id=163 data-nosnippet>163</a> <span class="kw">pub </span>flow_eval_interval: <span class="prelude-ty">Option</span><Duration>,
|
|
<a href=#164 id=164 data-nosnippet>164</a> <span class="doccomment">/// Typed schedule configuration, pre-parsed at task creation time.
|
|
<a href=#165 id=165 data-nosnippet>165</a> </span><span class="kw">pub </span>eval_schedule: <span class="prelude-ty">Option</span><EvalSchedule>,
|
|
<a href=#166 id=166 data-nosnippet>166</a>}
|
|
<a href=#167 id=167 data-nosnippet>167</a>
|
|
<a href=#168 id=168 data-nosnippet>168</a><span class="kw">fn </span>determine_query_type(query: <span class="kw-2">&</span>str, query_ctx: <span class="kw-2">&</span>QueryContextRef) -> <span class="prelude-ty">Result</span><QueryType, Error> {
|
|
<a href=#169 id=169 data-nosnippet>169</a> <span class="kw">let </span>is_tql = is_tql(query_ctx.sql_dialect(), query)
|
|
<a href=#170 id=170 data-nosnippet>170</a> .map_err(BoxedError::new)
|
|
<a href=#171 id=171 data-nosnippet>171</a> .context(ExternalSnafu)<span class="question-mark">?</span>;
|
|
<a href=#172 id=172 data-nosnippet>172</a> <span class="prelude-val">Ok</span>(<span class="kw">if </span>is_tql {
|
|
<a href=#173 id=173 data-nosnippet>173</a> QueryType::Tql
|
|
<a href=#174 id=174 data-nosnippet>174</a> } <span class="kw">else </span>{
|
|
<a href=#175 id=175 data-nosnippet>175</a> QueryType::Sql
|
|
<a href=#176 id=176 data-nosnippet>176</a> })
|
|
<a href=#177 id=177 data-nosnippet>177</a>}
|
|
<a href=#178 id=178 data-nosnippet>178</a>
|
|
<a href=#179 id=179 data-nosnippet>179</a><span class="kw">fn </span>is_merge_mode_last_non_null(options: <span class="kw-2">&</span>HashMap<String, String>) -> bool {
|
|
<a href=#180 id=180 data-nosnippet>180</a> options
|
|
<a href=#181 id=181 data-nosnippet>181</a> .get(MERGE_MODE_KEY)
|
|
<a href=#182 id=182 data-nosnippet>182</a> .map(|mode| mode.eq_ignore_ascii_case(<span class="string">"last_non_null"</span>))
|
|
<a href=#183 id=183 data-nosnippet>183</a> .unwrap_or(<span class="bool-val">false</span>)
|
|
<a href=#184 id=184 data-nosnippet>184</a>}
|
|
<a href=#185 id=185 data-nosnippet>185</a>
|
|
<a href=#186 id=186 data-nosnippet>186</a><span class="kw">fn </span>encode_insert_plan_request(
|
|
<a href=#187 id=187 data-nosnippet>187</a> insert_to: TableName,
|
|
<a href=#188 id=188 data-nosnippet>188</a> insert_input_plan: <span class="kw-2">&</span>LogicalPlan,
|
|
<a href=#189 id=189 data-nosnippet>189</a>) -> <span class="prelude-ty">Result</span><api::v1::QueryRequest, Error> {
|
|
<a href=#190 id=190 data-nosnippet>190</a> <span class="kw">let </span>message = DFLogicalSubstraitConvertor {}
|
|
<a href=#191 id=191 data-nosnippet>191</a> .encode(insert_input_plan, DefaultSerializer)
|
|
<a href=#192 id=192 data-nosnippet>192</a> .context(SubstraitEncodeLogicalPlanSnafu)<span class="question-mark">?</span>;
|
|
<a href=#193 id=193 data-nosnippet>193</a> <span class="prelude-val">Ok</span>(api::v1::QueryRequest {
|
|
<a href=#194 id=194 data-nosnippet>194</a> query: <span class="prelude-val">Some</span>(api::v1::query_request::Query::InsertIntoPlan(
|
|
<a href=#195 id=195 data-nosnippet>195</a> api::v1::InsertIntoPlan {
|
|
<a href=#196 id=196 data-nosnippet>196</a> table_name: <span class="prelude-val">Some</span>(insert_to),
|
|
<a href=#197 id=197 data-nosnippet>197</a> logical_plan: message.to_vec(),
|
|
<a href=#198 id=198 data-nosnippet>198</a> },
|
|
<a href=#199 id=199 data-nosnippet>199</a> )),
|
|
<a href=#200 id=200 data-nosnippet>200</a> })
|
|
<a href=#201 id=201 data-nosnippet>201</a>}
|
|
<a href=#202 id=202 data-nosnippet>202</a>
|
|
<a href=#203 id=203 data-nosnippet>203</a><span class="kw">fn </span>recovery_aggregate_input(plan: <span class="kw-2">&</span>LogicalPlan) -> <span class="prelude-ty">Result</span><LogicalPlan, Error> {
|
|
<a href=#204 id=204 data-nosnippet>204</a> <span class="kw">let </span>plan = <span class="kw">match </span>plan {
|
|
<a href=#205 id=205 data-nosnippet>205</a> LogicalPlan::Projection(projection) => projection.input.as_ref(),
|
|
<a href=#206 id=206 data-nosnippet>206</a> <span class="kw">_ </span>=> plan,
|
|
<a href=#207 id=207 data-nosnippet>207</a> };
|
|
<a href=#208 id=208 data-nosnippet>208</a> <span class="kw">let </span>LogicalPlan::Aggregate(aggregate) = plan <span class="kw">else </span>{
|
|
<a href=#209 id=209 data-nosnippet>209</a> <span class="kw">return </span>UnexpectedSnafu {
|
|
<a href=#210 id=210 data-nosnippet>210</a> reason: <span class="string">"Recovery timestamp projection did not find an aggregate"</span>.to_string(),
|
|
<a href=#211 id=211 data-nosnippet>211</a> }
|
|
<a href=#212 id=212 data-nosnippet>212</a> .fail();
|
|
<a href=#213 id=213 data-nosnippet>213</a> };
|
|
<a href=#214 id=214 data-nosnippet>214</a> <span class="prelude-val">Ok</span>(aggregate.input.as_ref().clone())
|
|
<a href=#215 id=215 data-nosnippet>215</a>}
|
|
<a href=#216 id=216 data-nosnippet>216</a>
|
|
<a href=#217 id=217 data-nosnippet>217</a><span class="kw">fn </span>capture_recovery_batch_windows(
|
|
<a href=#218 id=218 data-nosnippet>218</a> batch: <span class="kw-2">&</span>common_recordbatch::RecordBatch,
|
|
<a href=#219 id=219 data-nosnippet>219</a> time_window_expr: <span class="kw-2">&</span>TimeWindowExpr,
|
|
<a href=#220 id=220 data-nosnippet>220</a> windows: <span class="kw-2">&mut </span>BTreeSet<(Timestamp, Timestamp)>,
|
|
<a href=#221 id=221 data-nosnippet>221</a>) -> <span class="prelude-ty">Result</span><(), Error> {
|
|
<a href=#222 id=222 data-nosnippet>222</a> <span class="kw">if </span>batch.num_columns() != <span class="number">1 </span>{
|
|
<a href=#223 id=223 data-nosnippet>223</a> <span class="kw">return </span>UnexpectedSnafu {
|
|
<a href=#224 id=224 data-nosnippet>224</a> reason: <span class="macro">format!</span>(
|
|
<a href=#225 id=225 data-nosnippet>225</a> <span class="string">"Recovery timestamp projection returned {} columns instead of one"</span>,
|
|
<a href=#226 id=226 data-nosnippet>226</a> batch.num_columns()
|
|
<a href=#227 id=227 data-nosnippet>227</a> ),
|
|
<a href=#228 id=228 data-nosnippet>228</a> }
|
|
<a href=#229 id=229 data-nosnippet>229</a> .fail();
|
|
<a href=#230 id=230 data-nosnippet>230</a> }
|
|
<a href=#231 id=231 data-nosnippet>231</a> <span class="kw">let </span>values = Helper::try_into_vector(batch.column(<span class="number">0</span>).clone())
|
|
<a href=#232 id=232 data-nosnippet>232</a> .map_err(BoxedError::new)
|
|
<a href=#233 id=233 data-nosnippet>233</a> .context(ExternalSnafu)<span class="question-mark">?</span>;
|
|
<a href=#234 id=234 data-nosnippet>234</a> <span class="kw">for </span>index <span class="kw">in </span><span class="number">0</span>..values.len() {
|
|
<a href=#235 id=235 data-nosnippet>235</a> <span class="kw">let </span>timestamp = values.get(index).as_timestamp().context(UnexpectedSnafu {
|
|
<a href=#236 id=236 data-nosnippet>236</a> reason: <span class="string">"Recovery timestamp projection returned a null or non-timestamp value"
|
|
<a href=#237 id=237 data-nosnippet>237</a> </span>.to_string(),
|
|
<a href=#238 id=238 data-nosnippet>238</a> })<span class="question-mark">?</span>;
|
|
<a href=#239 id=239 data-nosnippet>239</a> <span class="kw">let </span>(start, end) = time_window_expr.eval(timestamp)<span class="question-mark">?</span>;
|
|
<a href=#240 id=240 data-nosnippet>240</a> <span class="kw">let </span>window = (
|
|
<a href=#241 id=241 data-nosnippet>241</a> start.context(UnexpectedSnafu {
|
|
<a href=#242 id=242 data-nosnippet>242</a> reason: <span class="string">"Recovery time-window expression returned no lower bound"</span>.to_string(),
|
|
<a href=#243 id=243 data-nosnippet>243</a> })<span class="question-mark">?</span>,
|
|
<a href=#244 id=244 data-nosnippet>244</a> end.context(UnexpectedSnafu {
|
|
<a href=#245 id=245 data-nosnippet>245</a> reason: <span class="string">"Recovery time-window expression returned no upper bound"</span>.to_string(),
|
|
<a href=#246 id=246 data-nosnippet>246</a> })<span class="question-mark">?</span>,
|
|
<a href=#247 id=247 data-nosnippet>247</a> );
|
|
<a href=#248 id=248 data-nosnippet>248</a> windows.insert(window);
|
|
<a href=#249 id=249 data-nosnippet>249</a> }
|
|
<a href=#250 id=250 data-nosnippet>250</a> <span class="prelude-val">Ok</span>(())
|
|
<a href=#251 id=251 data-nosnippet>251</a>}
|
|
<a href=#252 id=252 data-nosnippet>252</a>
|
|
<a href=#253 id=253 data-nosnippet>253</a><span class="kw">fn </span>format_insert_target_columns(plan: <span class="kw-2">&</span>LogicalPlan) -> String {
|
|
<a href=#254 id=254 data-nosnippet>254</a> plan.schema()
|
|
<a href=#255 id=255 data-nosnippet>255</a> .fields()
|
|
<a href=#256 id=256 data-nosnippet>256</a> .iter()
|
|
<a href=#257 id=257 data-nosnippet>257</a> .map(|field| quote_identifier(field.name()).to_string())
|
|
<a href=#258 id=258 data-nosnippet>258</a> .collect::<Vec<<span class="kw">_</span>>>()
|
|
<a href=#259 id=259 data-nosnippet>259</a> .join(<span class="string">", "</span>)
|
|
<a href=#260 id=260 data-nosnippet>260</a>}
|
|
<a href=#261 id=261 data-nosnippet>261</a>
|
|
<a href=#262 id=262 data-nosnippet>262</a><span class="doccomment">/// Owns a whole serialized execution round. It may be moved to an execution
|
|
<a href=#263 id=263 data-nosnippet>263</a>/// collaborator, so cancellation of the caller cannot release the round early.
|
|
<a href=#264 id=264 data-nosnippet>264</a></span><span class="kw">pub struct </span>BatchingExecutionGuard {
|
|
<a href=#265 id=265 data-nosnippet>265</a> _lock: OwnedMutexGuard<()>,
|
|
<a href=#266 id=266 data-nosnippet>266</a> restore: <span class="prelude-ty">Option</span><(Arc<RwLock<TaskState>>, QueryContextRef)>,
|
|
<a href=#267 id=267 data-nosnippet>267</a>}
|
|
<a href=#268 id=268 data-nosnippet>268</a>
|
|
<a href=#269 id=269 data-nosnippet>269</a><span class="kw">impl </span>BatchingExecutionGuard {
|
|
<a href=#270 id=270 data-nosnippet>270</a> <span class="kw">fn </span>new(lock: OwnedMutexGuard<()>) -> <span class="self">Self </span>{
|
|
<a href=#271 id=271 data-nosnippet>271</a> <span class="self">Self </span>{
|
|
<a href=#272 id=272 data-nosnippet>272</a> _lock: lock,
|
|
<a href=#273 id=273 data-nosnippet>273</a> restore: <span class="prelude-val">None</span>,
|
|
<a href=#274 id=274 data-nosnippet>274</a> }
|
|
<a href=#275 id=275 data-nosnippet>275</a> }
|
|
<a href=#276 id=276 data-nosnippet>276</a>
|
|
<a href=#277 id=277 data-nosnippet>277</a> <span class="kw">fn </span>restore_query_context(<span class="kw-2">&mut </span><span class="self">self</span>, state: Arc<RwLock<TaskState>>, old_ctx: QueryContextRef) {
|
|
<a href=#278 id=278 data-nosnippet>278</a> <span class="self">self</span>.restore = <span class="prelude-val">Some</span>((state, old_ctx));
|
|
<a href=#279 id=279 data-nosnippet>279</a> }
|
|
<a href=#280 id=280 data-nosnippet>280</a>}
|
|
<a href=#281 id=281 data-nosnippet>281</a>
|
|
<a href=#282 id=282 data-nosnippet>282</a><span class="kw">impl </span>Drop <span class="kw">for </span>BatchingExecutionGuard {
|
|
<a href=#283 id=283 data-nosnippet>283</a> <span class="kw">fn </span>drop(<span class="kw-2">&mut </span><span class="self">self</span>) {
|
|
<a href=#284 id=284 data-nosnippet>284</a> <span class="kw">if let </span><span class="prelude-val">Some</span>((state, old_ctx)) = <span class="self">self</span>.restore.take()
|
|
<a href=#285 id=285 data-nosnippet>285</a> && <span class="kw">let </span><span class="prelude-val">Ok</span>(<span class="kw-2">mut </span>state) = state.write()
|
|
<a href=#286 id=286 data-nosnippet>286</a> {
|
|
<a href=#287 id=287 data-nosnippet>287</a> state.query_ctx = old_ctx;
|
|
<a href=#288 id=288 data-nosnippet>288</a> }
|
|
<a href=#289 id=289 data-nosnippet>289</a> }
|
|
<a href=#290 id=290 data-nosnippet>290</a>}
|
|
<a href=#291 id=291 data-nosnippet>291</a>
|
|
<a href=#292 id=292 data-nosnippet>292</a><span class="attr">#[derive(Clone)]
|
|
<a href=#293 id=293 data-nosnippet>293</a></span><span class="kw">pub struct </span>BatchingTask {
|
|
<a href=#294 id=294 data-nosnippet>294</a> <span class="kw">pub </span>config: Arc<TaskConfig>,
|
|
<a href=#295 id=295 data-nosnippet>295</a> <span class="kw">pub </span>state: Arc<RwLock<TaskState>>,
|
|
<a href=#296 id=296 data-nosnippet>296</a> <span class="doccomment">/// Serializes plan generation, execution, checkpoint advancement, and dirty
|
|
<a href=#297 id=297 data-nosnippet>297</a> /// window restoration for this flow. Without this, a manual flush and the
|
|
<a href=#298 id=298 data-nosnippet>298</a> /// background loop can process the same checkpoint range concurrently.
|
|
<a href=#299 id=299 data-nosnippet>299</a> </span>execution_lock: Arc<Mutex<()>>,
|
|
<a href=#300 id=300 data-nosnippet>300</a> execution: <span class="prelude-ty">Option</span><Arc<<span class="kw">dyn </span><span class="kw">crate</span>::BatchingExecution>>,
|
|
<a href=#301 id=301 data-nosnippet>301</a>}
|
|
<a href=#302 id=302 data-nosnippet>302</a>
|
|
<a href=#303 id=303 data-nosnippet>303</a><span class="doccomment">/// Arguments for creating batching task
|
|
<a href=#304 id=304 data-nosnippet>304</a></span><span class="kw">pub struct </span>TaskArgs<<span class="lifetime">'a</span>> {
|
|
<a href=#305 id=305 data-nosnippet>305</a> <span class="kw">pub </span>flow_id: FlowId,
|
|
<a href=#306 id=306 data-nosnippet>306</a> <span class="kw">pub </span>query: <span class="kw-2">&</span><span class="lifetime">'a </span>str,
|
|
<a href=#307 id=307 data-nosnippet>307</a> <span class="kw">pub </span>plan: LogicalPlan,
|
|
<a href=#308 id=308 data-nosnippet>308</a> <span class="kw">pub </span>time_window_expr: <span class="prelude-ty">Option</span><TimeWindowExpr>,
|
|
<a href=#309 id=309 data-nosnippet>309</a> <span class="kw">pub </span>expire_after: <span class="prelude-ty">Option</span><i64>,
|
|
<a href=#310 id=310 data-nosnippet>310</a> <span class="kw">pub </span>sink_table_name: [String; <span class="number">3</span>],
|
|
<a href=#311 id=311 data-nosnippet>311</a> <span class="kw">pub </span>source_table_names: Vec<[String; <span class="number">3</span>]>,
|
|
<a href=#312 id=312 data-nosnippet>312</a> <span class="kw">pub </span>query_ctx: QueryContextRef,
|
|
<a href=#313 id=313 data-nosnippet>313</a> <span class="kw">pub </span>catalog_manager: CatalogManagerRef,
|
|
<a href=#314 id=314 data-nosnippet>314</a> <span class="kw">pub </span>shutdown_rx: oneshot::Receiver<()>,
|
|
<a href=#315 id=315 data-nosnippet>315</a> <span class="kw">pub </span>batch_opts: Arc<BatchingModeOptions>,
|
|
<a href=#316 id=316 data-nosnippet>316</a> <span class="kw">pub </span>flow_eval_interval: <span class="prelude-ty">Option</span><Duration>,
|
|
<a href=#317 id=317 data-nosnippet>317</a> <span class="doccomment">/// Typed schedule configuration pre-parsed from `CreateFlowArgs`.
|
|
<a href=#318 id=318 data-nosnippet>318</a> </span><span class="kw">pub </span>eval_schedule: <span class="prelude-ty">Option</span><EvalSchedule>,
|
|
<a href=#319 id=319 data-nosnippet>319</a>}
|
|
<a href=#320 id=320 data-nosnippet>320</a>
|
|
<a href=#321 id=321 data-nosnippet>321</a><span class="kw">pub struct </span>PlanInfo {
|
|
<a href=#322 id=322 data-nosnippet>322</a> <span class="kw">pub </span>plan: LogicalPlan,
|
|
<a href=#323 id=323 data-nosnippet>323</a> <span class="kw">pub </span>dirty_restore: DirtyRestore,
|
|
<a href=#324 id=324 data-nosnippet>324</a> <span class="kw">pub </span>coverage: QueryCoverage,
|
|
<a href=#325 id=325 data-nosnippet>325</a>}
|
|
<a href=#326 id=326 data-nosnippet>326</a>
|
|
<a href=#327 id=327 data-nosnippet>327</a><span class="attr">#[derive(Clone)]
|
|
<a href=#328 id=328 data-nosnippet>328</a></span><span class="kw">pub enum </span>QueryCoverage {
|
|
<a href=#329 id=329 data-nosnippet>329</a> <span class="doccomment">/// Explicit full-query snapshot coverage, e.g. TQL or evaluation-interval
|
|
<a href=#330 id=330 data-nosnippet>330</a> /// SQL flows whose plan shape cannot be safely dirty-window pruned. This
|
|
<a href=#331 id=331 data-nosnippet>331</a> /// must not be used as an implicit recovery path for scoped repair or an
|
|
<a href=#332 id=332 data-nosnippet>332</a> /// unsafe incremental rewrite fallback.
|
|
<a href=#333 id=333 data-nosnippet>333</a> </span>UnfilteredFull,
|
|
<a href=#334 id=334 data-nosnippet>334</a> <span class="doccomment">/// Scoped full-snapshot repair over the current dirty windows. A successful
|
|
<a href=#335 id=335 data-nosnippet>335</a> /// result may start a fenced repair if new dirty windows appeared meanwhile.
|
|
<a href=#336 id=336 data-nosnippet>336</a> </span>ScopedBaseRepair,
|
|
<a href=#337 id=337 data-nosnippet>337</a> <span class="doccomment">/// A chunk of windows being repaired under the frozen high-watermark `H`.
|
|
<a href=#338 id=338 data-nosnippet>338</a> /// The `high` map is sent as snapshot read bounds and must be matched by
|
|
<a href=#339 id=339 data-nosnippet>339</a> /// the returned terminal watermarks before checkpoints can advance.
|
|
<a href=#340 id=340 data-nosnippet>340</a> </span>FencedRepairChunk { high: BTreeMap<u64, u64> },
|
|
<a href=#341 id=341 data-nosnippet>341</a> <span class="doccomment">/// Incremental delta query over `(checkpoint, scan-open snapshot]`.
|
|
<a href=#342 id=342 data-nosnippet>342</a> </span>IncrementalDelta,
|
|
<a href=#343 id=343 data-nosnippet>343</a>}
|
|
<a href=#344 id=344 data-nosnippet>344</a>
|
|
<a href=#345 id=345 data-nosnippet>345</a><span class="kw">impl </span>QueryCoverage {
|
|
<a href=#346 id=346 data-nosnippet>346</a> <span class="doccomment">/// Whether this query should use incremental scan extensions and
|
|
<a href=#347 id=347 data-nosnippet>347</a> /// incremental checkpoint advancement rules.
|
|
<a href=#348 id=348 data-nosnippet>348</a> </span><span class="kw">fn </span>is_incremental_delta(<span class="kw-2">&</span><span class="self">self</span>) -> bool {
|
|
<a href=#349 id=349 data-nosnippet>349</a> <span class="macro">matches!</span>(<span class="self">self</span>, <span class="self">Self</span>::IncrementalDelta)
|
|
<a href=#350 id=350 data-nosnippet>350</a> }
|
|
<a href=#351 id=351 data-nosnippet>351</a>
|
|
<a href=#352 id=352 data-nosnippet>352</a> <span class="doccomment">/// Snapshot upper bounds requested from the storage layer. Only fenced
|
|
<a href=#353 id=353 data-nosnippet>353</a> /// repair chunks carry bounds; all other coverage relies on normal scans.
|
|
<a href=#354 id=354 data-nosnippet>354</a> </span><span class="kw">fn </span>snapshot_seqs(<span class="kw-2">&</span><span class="self">self</span>) -> HashMap<u64, u64> {
|
|
<a href=#355 id=355 data-nosnippet>355</a> <span class="kw">match </span><span class="self">self </span>{
|
|
<a href=#356 id=356 data-nosnippet>356</a> <span class="self">Self</span>::FencedRepairChunk { high } => high.iter().map(|(k, v)| (<span class="kw-2">*</span>k, <span class="kw-2">*</span>v)).collect(),
|
|
<a href=#357 id=357 data-nosnippet>357</a> <span class="kw">_ </span>=> HashMap::new(),
|
|
<a href=#358 id=358 data-nosnippet>358</a> }
|
|
<a href=#359 id=359 data-nosnippet>359</a> }
|
|
<a href=#360 id=360 data-nosnippet>360</a>}
|
|
<a href=#361 id=361 data-nosnippet>361</a>
|
|
<a href=#362 id=362 data-nosnippet>362</a><span class="kw">pub enum </span>DirtyRestore {
|
|
<a href=#363 id=363 data-nosnippet>363</a> <span class="doccomment">/// The query was scoped to dirty time ranges; restore those ranges if the
|
|
<a href=#364 id=364 data-nosnippet>364</a> /// run fails.
|
|
<a href=#365 id=365 data-nosnippet>365</a> </span>Scoped(FilterExprInfo),
|
|
<a href=#366 id=366 data-nosnippet>366</a> <span class="doccomment">/// The query could not be scoped to dirty time ranges, so the dirty-window
|
|
<a href=#367 id=367 data-nosnippet>367</a> /// state is only a dirty signal. Restore the consumed signal if the full
|
|
<a href=#368 id=368 data-nosnippet>368</a> /// run fails.
|
|
<a href=#369 id=369 data-nosnippet>369</a> ///
|
|
<a href=#370 id=370 data-nosnippet>370</a> /// TODO(discord9): Full-query runs only need a dirty bool flag. Refactor
|
|
<a href=#371 id=371 data-nosnippet>371</a> /// the unscoped path to stop reusing `DirtyTimeWindows` for this signal.
|
|
<a href=#372 id=372 data-nosnippet>372</a> </span>Unscoped(DirtyTimeWindows),
|
|
<a href=#373 id=373 data-nosnippet>373</a>}
|
|
<a href=#374 id=374 data-nosnippet>374</a>
|
|
<a href=#375 id=375 data-nosnippet>375</a><span class="kw">pub struct </span>ExecuteOnceOutcome {
|
|
<a href=#376 id=376 data-nosnippet>376</a> <span class="kw">pub </span>new_query: <span class="prelude-ty">Option</span><PlanInfo>,
|
|
<a href=#377 id=377 data-nosnippet>377</a> <span class="doccomment">/// Execution result of the generated insert plan.
|
|
<a href=#378 id=378 data-nosnippet>378</a> ///
|
|
<a href=#379 id=379 data-nosnippet>379</a> /// `Ok(Some((affected_rows, elapsed)))` means a query was executed.
|
|
<a href=#380 id=380 data-nosnippet>380</a> /// `Ok(None)` means no query was generated because there was no dirty signal.
|
|
<a href=#381 id=381 data-nosnippet>381</a> /// `Err(_)` means plan generation or execution failed.
|
|
<a href=#382 id=382 data-nosnippet>382</a> </span><span class="kw">pub </span>result: <span class="prelude-ty">Result</span><<span class="prelude-ty">Option</span><(usize, Duration)>, Error>,
|
|
<a href=#383 id=383 data-nosnippet>383</a>}
|
|
<a href=#384 id=384 data-nosnippet>384</a>
|
|
<a href=#385 id=385 data-nosnippet>385</a><span class="kw">impl </span>BatchingTask {
|
|
<a href=#386 id=386 data-nosnippet>386</a> <span class="attr">#[allow(clippy::too_many_arguments)]
|
|
<a href=#387 id=387 data-nosnippet>387</a> </span><span class="kw">pub fn </span>try_new(args: TaskArgs<<span class="lifetime">'_</span>>) -> <span class="prelude-ty">Result</span><<span class="self">Self</span>, Error> {
|
|
<a href=#388 id=388 data-nosnippet>388</a> <span class="self">Self</span>::try_new_with_exact_sequence_range_required(args, <span class="bool-val">false</span>)
|
|
<a href=#389 id=389 data-nosnippet>389</a> }
|
|
<a href=#390 id=390 data-nosnippet>390</a>
|
|
<a href=#391 id=391 data-nosnippet>391</a> <span class="kw">pub fn </span>try_new_with_exact_sequence_range_required(
|
|
<a href=#392 id=392 data-nosnippet>392</a> TaskArgs {
|
|
<a href=#393 id=393 data-nosnippet>393</a> flow_id,
|
|
<a href=#394 id=394 data-nosnippet>394</a> query,
|
|
<a href=#395 id=395 data-nosnippet>395</a> plan,
|
|
<a href=#396 id=396 data-nosnippet>396</a> time_window_expr,
|
|
<a href=#397 id=397 data-nosnippet>397</a> expire_after,
|
|
<a href=#398 id=398 data-nosnippet>398</a> sink_table_name,
|
|
<a href=#399 id=399 data-nosnippet>399</a> source_table_names,
|
|
<a href=#400 id=400 data-nosnippet>400</a> query_ctx,
|
|
<a href=#401 id=401 data-nosnippet>401</a> catalog_manager,
|
|
<a href=#402 id=402 data-nosnippet>402</a> shutdown_rx,
|
|
<a href=#403 id=403 data-nosnippet>403</a> batch_opts,
|
|
<a href=#404 id=404 data-nosnippet>404</a> flow_eval_interval,
|
|
<a href=#405 id=405 data-nosnippet>405</a> eval_schedule,
|
|
<a href=#406 id=406 data-nosnippet>406</a> }: TaskArgs<<span class="lifetime">'_</span>>,
|
|
<a href=#407 id=407 data-nosnippet>407</a> exact_sequence_range_required: bool,
|
|
<a href=#408 id=408 data-nosnippet>408</a> ) -> <span class="prelude-ty">Result</span><<span class="self">Self</span>, Error> {
|
|
<a href=#409 id=409 data-nosnippet>409</a> <span class="kw">let </span><span class="kw-2">mut </span>state = TaskState::with_dirty_time_windows(
|
|
<a href=#410 id=410 data-nosnippet>410</a> query_ctx.clone(),
|
|
<a href=#411 id=411 data-nosnippet>411</a> shutdown_rx,
|
|
<a href=#412 id=412 data-nosnippet>412</a> DirtyTimeWindows::new(
|
|
<a href=#413 id=413 data-nosnippet>413</a> batch_opts.experimental_max_filter_num_per_query,
|
|
<a href=#414 id=414 data-nosnippet>414</a> batch_opts.experimental_time_window_merge_threshold,
|
|
<a href=#415 id=415 data-nosnippet>415</a> ),
|
|
<a href=#416 id=416 data-nosnippet>416</a> );
|
|
<a href=#417 id=417 data-nosnippet>417</a> <span class="kw">if </span>!batch_opts.experimental_enable_incremental_read {
|
|
<a href=#418 id=418 data-nosnippet>418</a> state.disable_incremental();
|
|
<a href=#419 id=419 data-nosnippet>419</a> }
|
|
<a href=#420 id=420 data-nosnippet>420</a>
|
|
<a href=#421 id=421 data-nosnippet>421</a> <span class="prelude-val">Ok</span>(<span class="self">Self </span>{
|
|
<a href=#422 id=422 data-nosnippet>422</a> config: Arc::new(TaskConfig {
|
|
<a href=#423 id=423 data-nosnippet>423</a> flow_id,
|
|
<a href=#424 id=424 data-nosnippet>424</a> query: query.to_string(),
|
|
<a href=#425 id=425 data-nosnippet>425</a> time_window_expr,
|
|
<a href=#426 id=426 data-nosnippet>426</a> expire_after,
|
|
<a href=#427 id=427 data-nosnippet>427</a> sink_table_name,
|
|
<a href=#428 id=428 data-nosnippet>428</a> source_table_names: source_table_names.into_iter().collect(),
|
|
<a href=#429 id=429 data-nosnippet>429</a> catalog_manager,
|
|
<a href=#430 id=430 data-nosnippet>430</a> output_schema: plan.schema().clone(),
|
|
<a href=#431 id=431 data-nosnippet>431</a> query_type: determine_query_type(query, <span class="kw-2">&</span>query_ctx)<span class="question-mark">?</span>,
|
|
<a href=#432 id=432 data-nosnippet>432</a> exact_sequence_range_required,
|
|
<a href=#433 id=433 data-nosnippet>433</a> batch_opts,
|
|
<a href=#434 id=434 data-nosnippet>434</a> flow_eval_interval,
|
|
<a href=#435 id=435 data-nosnippet>435</a> eval_schedule,
|
|
<a href=#436 id=436 data-nosnippet>436</a> }),
|
|
<a href=#437 id=437 data-nosnippet>437</a> state: Arc::new(RwLock::new(state)),
|
|
<a href=#438 id=438 data-nosnippet>438</a> execution_lock: Arc::new(Mutex::new(())),
|
|
<a href=#439 id=439 data-nosnippet>439</a> execution: <span class="prelude-val">None</span>,
|
|
<a href=#440 id=440 data-nosnippet>440</a> })
|
|
<a href=#441 id=441 data-nosnippet>441</a> }
|
|
<a href=#442 id=442 data-nosnippet>442</a>
|
|
<a href=#443 id=443 data-nosnippet>443</a> <span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">fn </span>with_execution(
|
|
<a href=#444 id=444 data-nosnippet>444</a> <span class="kw-2">mut </span><span class="self">self</span>,
|
|
<a href=#445 id=445 data-nosnippet>445</a> execution: <span class="prelude-ty">Option</span><Arc<<span class="kw">dyn </span><span class="kw">crate</span>::BatchingExecution>>,
|
|
<a href=#446 id=446 data-nosnippet>446</a> ) -> <span class="self">Self </span>{
|
|
<a href=#447 id=447 data-nosnippet>447</a> <span class="self">self</span>.execution = execution;
|
|
<a href=#448 id=448 data-nosnippet>448</a> <span class="self">self
|
|
<a href=#449 id=449 data-nosnippet>449</a> </span>}
|
|
<a href=#450 id=450 data-nosnippet>450</a>
|
|
<a href=#451 id=451 data-nosnippet>451</a> <span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">fn </span>stop_execution(<span class="kw-2">&</span><span class="self">self</span>) {
|
|
<a href=#452 id=452 data-nosnippet>452</a> <span class="kw">if let </span><span class="prelude-val">Some</span>(execution) = <span class="kw-2">&</span><span class="self">self</span>.execution {
|
|
<a href=#453 id=453 data-nosnippet>453</a> execution.stop();
|
|
<a href=#454 id=454 data-nosnippet>454</a> }
|
|
<a href=#455 id=455 data-nosnippet>455</a> }
|
|
<a href=#456 id=456 data-nosnippet>456</a>
|
|
<a href=#457 id=457 data-nosnippet>457</a> <span class="kw">pub fn </span>last_execution_time_millis(<span class="kw-2">&</span><span class="self">self</span>) -> <span class="prelude-ty">Option</span><i64> {
|
|
<a href=#458 id=458 data-nosnippet>458</a> <span class="self">self</span>.state.read().unwrap().last_execution_time_millis()
|
|
<a href=#459 id=459 data-nosnippet>459</a> }
|
|
<a href=#460 id=460 data-nosnippet>460</a>
|
|
<a href=#461 id=461 data-nosnippet>461</a> <span class="kw">pub fn </span>start_time_millis(<span class="kw-2">&</span><span class="self">self</span>) -> <span class="prelude-ty">Option</span><i64> {
|
|
<a href=#462 id=462 data-nosnippet>462</a> <span class="self">self</span>.state.read().unwrap().start_time_millis()
|
|
<a href=#463 id=463 data-nosnippet>463</a> }
|
|
<a href=#464 id=464 data-nosnippet>464</a>
|
|
<a href=#465 id=465 data-nosnippet>465</a> <span class="doccomment">/// Collect flow-related extensions from the task's query context that should be
|
|
<a href=#466 id=466 data-nosnippet>466</a> /// forwarded to the frontend (e.g. scheduled time).
|
|
<a href=#467 id=467 data-nosnippet>467</a> </span><span class="kw">fn </span>frontend_extensions(<span class="kw-2">&</span><span class="self">self</span>) -> HashMap<String, String> {
|
|
<a href=#468 id=468 data-nosnippet>468</a> <span class="kw">let </span>ctx = <span class="self">self</span>.state.read().unwrap();
|
|
<a href=#469 id=469 data-nosnippet>469</a> <span class="kw">let </span>all = ctx.query_ctx.extensions();
|
|
<a href=#470 id=470 data-nosnippet>470</a> <span class="kw">let </span><span class="kw-2">mut </span>flow_exts = HashMap::new();
|
|
<a href=#471 id=471 data-nosnippet>471</a> <span class="comment">// Propagate the scheduled time extension if present so that frontend
|
|
<a href=#472 id=472 data-nosnippet>472</a> // execution can use the same logical time.
|
|
<a href=#473 id=473 data-nosnippet>473</a> </span><span class="kw">if let </span><span class="prelude-val">Some</span>(v) = all.get(query::options::FLOW_SCHEDULED_TIME_MILLIS) {
|
|
<a href=#474 id=474 data-nosnippet>474</a> flow_exts.insert(
|
|
<a href=#475 id=475 data-nosnippet>475</a> query::options::FLOW_SCHEDULED_TIME_MILLIS.to_string(),
|
|
<a href=#476 id=476 data-nosnippet>476</a> v.clone(),
|
|
<a href=#477 id=477 data-nosnippet>477</a> );
|
|
<a href=#478 id=478 data-nosnippet>478</a> }
|
|
<a href=#479 id=479 data-nosnippet>479</a> flow_exts
|
|
<a href=#480 id=480 data-nosnippet>480</a> }
|
|
<a href=#481 id=481 data-nosnippet>481</a>
|
|
<a href=#482 id=482 data-nosnippet>482</a> <span class="doccomment">/// mark time window range (now - expire_after, now) as dirty (or (0, now) if expire_after not set)
|
|
<a href=#483 id=483 data-nosnippet>483</a> ///
|
|
<a href=#484 id=484 data-nosnippet>484</a> /// useful for flush_flow to flush dirty time windows range
|
|
<a href=#485 id=485 data-nosnippet>485</a> </span><span class="kw">pub fn </span>mark_all_windows_as_dirty(<span class="kw-2">&</span><span class="self">self</span>) -> <span class="prelude-ty">Result</span><(), Error> {
|
|
<a href=#486 id=486 data-nosnippet>486</a> <span class="kw">let </span>now = SystemTime::now();
|
|
<a href=#487 id=487 data-nosnippet>487</a> <span class="kw">let </span>now = Timestamp::new_second(
|
|
<a href=#488 id=488 data-nosnippet>488</a> now.duration_since(UNIX_EPOCH)
|
|
<a href=#489 id=489 data-nosnippet>489</a> .expect(<span class="string">"Time went backwards"</span>)
|
|
<a href=#490 id=490 data-nosnippet>490</a> .as_secs() <span class="kw">as _</span>,
|
|
<a href=#491 id=491 data-nosnippet>491</a> );
|
|
<a href=#492 id=492 data-nosnippet>492</a> <span class="kw">let </span>lower_bound = <span class="self">self
|
|
<a href=#493 id=493 data-nosnippet>493</a> </span>.config
|
|
<a href=#494 id=494 data-nosnippet>494</a> .expire_after
|
|
<a href=#495 id=495 data-nosnippet>495</a> .map(|e| now.sub_duration(Duration::from_secs(e <span class="kw">as _</span>)))
|
|
<a href=#496 id=496 data-nosnippet>496</a> .transpose()
|
|
<a href=#497 id=497 data-nosnippet>497</a> .map_err(BoxedError::new)
|
|
<a href=#498 id=498 data-nosnippet>498</a> .context(ExternalSnafu)<span class="question-mark">?
|
|
<a href=#499 id=499 data-nosnippet>499</a> </span>.unwrap_or(Timestamp::new_second(<span class="number">0</span>));
|
|
<a href=#500 id=500 data-nosnippet>500</a> <span class="macro">debug!</span>(
|
|
<a href=#501 id=501 data-nosnippet>501</a> <span class="string">"Flow {} mark range ({:?}, {:?}) as dirty"</span>,
|
|
<a href=#502 id=502 data-nosnippet>502</a> <span class="self">self</span>.config.flow_id, lower_bound, now
|
|
<a href=#503 id=503 data-nosnippet>503</a> );
|
|
<a href=#504 id=504 data-nosnippet>504</a> <span class="self">self</span>.state
|
|
<a href=#505 id=505 data-nosnippet>505</a> .write()
|
|
<a href=#506 id=506 data-nosnippet>506</a> .unwrap()
|
|
<a href=#507 id=507 data-nosnippet>507</a> .dirty_time_windows
|
|
<a href=#508 id=508 data-nosnippet>508</a> .add_window(lower_bound, <span class="prelude-val">Some</span>(now));
|
|
<a href=#509 id=509 data-nosnippet>509</a> <span class="prelude-val">Ok</span>(())
|
|
<a href=#510 id=510 data-nosnippet>510</a> }
|
|
<a href=#511 id=511 data-nosnippet>511</a>
|
|
<a href=#512 id=512 data-nosnippet>512</a> <span class="doccomment">/// Create the sink table if needed, then return it.
|
|
<a href=#513 id=513 data-nosnippet>513</a> </span><span class="kw">pub async fn </span>check_or_create_sink_table(
|
|
<a href=#514 id=514 data-nosnippet>514</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#515 id=515 data-nosnippet>515</a> engine: <span class="kw-2">&</span>QueryEngineRef,
|
|
<a href=#516 id=516 data-nosnippet>516</a> frontend_client: <span class="kw-2">&</span>Arc<FrontendClient>,
|
|
<a href=#517 id=517 data-nosnippet>517</a> ) -> <span class="prelude-ty">Result</span><TableRef, Error> {
|
|
<a href=#518 id=518 data-nosnippet>518</a> <span class="kw">if </span>!<span class="self">self</span>.is_table_exist(<span class="kw-2">&</span><span class="self">self</span>.config.sink_table_name).<span class="kw">await</span><span class="question-mark">? </span>{
|
|
<a href=#519 id=519 data-nosnippet>519</a> <span class="kw">let </span>create_table = <span class="self">self</span>.gen_create_table_expr(engine.clone()).<span class="kw">await</span><span class="question-mark">?</span>;
|
|
<a href=#520 id=520 data-nosnippet>520</a> <span class="macro">info!</span>(
|
|
<a href=#521 id=521 data-nosnippet>521</a> <span class="string">"Try creating sink table(if not exists) with expr: {:?}"</span>,
|
|
<a href=#522 id=522 data-nosnippet>522</a> create_table
|
|
<a href=#523 id=523 data-nosnippet>523</a> );
|
|
<a href=#524 id=524 data-nosnippet>524</a> <span class="self">self</span>.create_table(frontend_client, create_table).<span class="kw">await</span><span class="question-mark">?</span>;
|
|
<a href=#525 id=525 data-nosnippet>525</a> <span class="macro">info!</span>(
|
|
<a href=#526 id=526 data-nosnippet>526</a> <span class="string">"Sink table {}(if not exists) created"</span>,
|
|
<a href=#527 id=527 data-nosnippet>527</a> <span class="self">self</span>.config.sink_table_name.join(<span class="string">"."</span>)
|
|
<a href=#528 id=528 data-nosnippet>528</a> );
|
|
<a href=#529 id=529 data-nosnippet>529</a> }
|
|
<a href=#530 id=530 data-nosnippet>530</a> <span class="kw">let </span>(table, <span class="kw">_</span>) = get_table_info_df_schema(
|
|
<a href=#531 id=531 data-nosnippet>531</a> <span class="self">self</span>.config.catalog_manager.clone(),
|
|
<a href=#532 id=532 data-nosnippet>532</a> <span class="self">self</span>.config.sink_table_name.clone(),
|
|
<a href=#533 id=533 data-nosnippet>533</a> )
|
|
<a href=#534 id=534 data-nosnippet>534</a> .<span class="kw">await</span><span class="question-mark">?</span>;
|
|
<a href=#535 id=535 data-nosnippet>535</a> <span class="prelude-val">Ok</span>(table)
|
|
<a href=#536 id=536 data-nosnippet>536</a> }
|
|
<a href=#537 id=537 data-nosnippet>537</a>
|
|
<a href=#538 id=538 data-nosnippet>538</a> <span class="doccomment">/// Returns a forked snapshot of the task query context for extension-owned planning.
|
|
<a href=#539 id=539 data-nosnippet>539</a> </span><span class="kw">pub fn </span>query_context_snapshot(<span class="kw-2">&</span><span class="self">self</span>) -> QueryContextRef {
|
|
<a href=#540 id=540 data-nosnippet>540</a> <span class="kw">let </span>query_ctx = <span class="self">self</span>.state.read().unwrap().query_ctx.clone();
|
|
<a href=#541 id=541 data-nosnippet>541</a> Arc::new(query_ctx.fork())
|
|
<a href=#542 id=542 data-nosnippet>542</a> }
|
|
<a href=#543 id=543 data-nosnippet>543</a>
|
|
<a href=#544 id=544 data-nosnippet>544</a> <span class="doccomment">/// Returns the retention lower bound aligned to this task's time window.
|
|
<a href=#545 id=545 data-nosnippet>545</a> </span><span class="kw">pub fn </span>recovery_retention_lower_bound(<span class="kw-2">&</span><span class="self">self</span>) -> <span class="prelude-ty">Result</span><<span class="prelude-ty">Option</span><Timestamp>, Error> {
|
|
<a href=#546 id=546 data-nosnippet>546</a> <span class="kw">let </span><span class="prelude-val">Some</span>(expire_after) = <span class="self">self</span>.config.expire_after <span class="kw">else </span>{
|
|
<a href=#547 id=547 data-nosnippet>547</a> <span class="kw">return </span><span class="prelude-val">Ok</span>(<span class="prelude-val">None</span>);
|
|
<a href=#548 id=548 data-nosnippet>548</a> };
|
|
<a href=#549 id=549 data-nosnippet>549</a> <span class="kw">let </span>expire_after = u64::try_from(expire_after).map_err(|<span class="kw">_</span>| {
|
|
<a href=#550 id=550 data-nosnippet>550</a> UnexpectedSnafu {
|
|
<a href=#551 id=551 data-nosnippet>551</a> reason: <span class="macro">format!</span>(
|
|
<a href=#552 id=552 data-nosnippet>552</a> <span class="string">"Flow {} has negative expire_after {expire_after}"</span>,
|
|
<a href=#553 id=553 data-nosnippet>553</a> <span class="self">self</span>.config.flow_id
|
|
<a href=#554 id=554 data-nosnippet>554</a> ),
|
|
<a href=#555 id=555 data-nosnippet>555</a> }
|
|
<a href=#556 id=556 data-nosnippet>556</a> .build()
|
|
<a href=#557 id=557 data-nosnippet>557</a> })<span class="question-mark">?</span>;
|
|
<a href=#558 id=558 data-nosnippet>558</a> <span class="kw">let </span>now = Timestamp::new_second(
|
|
<a href=#559 id=559 data-nosnippet>559</a> SystemTime::now()
|
|
<a href=#560 id=560 data-nosnippet>560</a> .duration_since(UNIX_EPOCH)
|
|
<a href=#561 id=561 data-nosnippet>561</a> .map_err(|err| {
|
|
<a href=#562 id=562 data-nosnippet>562</a> UnexpectedSnafu {
|
|
<a href=#563 id=563 data-nosnippet>563</a> reason: <span class="macro">format!</span>(<span class="string">"Failed to read recovery wall clock: {err}"</span>),
|
|
<a href=#564 id=564 data-nosnippet>564</a> }
|
|
<a href=#565 id=565 data-nosnippet>565</a> .build()
|
|
<a href=#566 id=566 data-nosnippet>566</a> })<span class="question-mark">?
|
|
<a href=#567 id=567 data-nosnippet>567</a> </span>.as_secs() <span class="kw">as </span>i64,
|
|
<a href=#568 id=568 data-nosnippet>568</a> );
|
|
<a href=#569 id=569 data-nosnippet>569</a> <span class="kw">let </span>lower = now
|
|
<a href=#570 id=570 data-nosnippet>570</a> .sub_duration(Duration::from_secs(expire_after))
|
|
<a href=#571 id=571 data-nosnippet>571</a> .map_err(BoxedError::new)
|
|
<a href=#572 id=572 data-nosnippet>572</a> .context(ExternalSnafu)<span class="question-mark">?</span>;
|
|
<a href=#573 id=573 data-nosnippet>573</a> <span class="self">self</span>.config
|
|
<a href=#574 id=574 data-nosnippet>574</a> .time_window_expr
|
|
<a href=#575 id=575 data-nosnippet>575</a> .as_ref()
|
|
<a href=#576 id=576 data-nosnippet>576</a> .context(UnexpectedSnafu {
|
|
<a href=#577 id=577 data-nosnippet>577</a> reason: <span class="string">"Recovery expiry requires a time-window expression"</span>.to_string(),
|
|
<a href=#578 id=578 data-nosnippet>578</a> })<span class="question-mark">?
|
|
<a href=#579 id=579 data-nosnippet>579</a> </span>.eval(lower)<span class="question-mark">?
|
|
<a href=#580 id=580 data-nosnippet>580</a> </span>.<span class="number">0
|
|
<a href=#581 id=581 data-nosnippet>581</a> </span>.context(UnexpectedSnafu {
|
|
<a href=#582 id=582 data-nosnippet>582</a> reason: <span class="string">"Recovery expiry time-window expression returned no lower bound"
|
|
<a href=#583 id=583 data-nosnippet>583</a> </span>.to_string(),
|
|
<a href=#584 id=584 data-nosnippet>584</a> })
|
|
<a href=#585 id=585 data-nosnippet>585</a> .map(<span class="prelude-val">Some</span>)
|
|
<a href=#586 id=586 data-nosnippet>586</a> }
|
|
<a href=#587 id=587 data-nosnippet>587</a>
|
|
<a href=#588 id=588 data-nosnippet>588</a> <span class="doccomment">/// Verifies that recovery can still read all source data required by its retained scope.
|
|
<a href=#589 id=589 data-nosnippet>589</a> </span><span class="kw">pub async fn </span>validate_recovery_retention(
|
|
<a href=#590 id=590 data-nosnippet>590</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#591 id=591 data-nosnippet>591</a> retention_lower: <span class="prelude-ty">Option</span><Timestamp>,
|
|
<a href=#592 id=592 data-nosnippet>592</a> windows: <span class="kw-2">&</span>[(Timestamp, Timestamp)],
|
|
<a href=#593 id=593 data-nosnippet>593</a> ) -> <span class="prelude-ty">Result</span><(), Error> {
|
|
<a href=#594 id=594 data-nosnippet>594</a> <span class="kw">for </span>name <span class="kw">in </span><span class="kw-2">&</span><span class="self">self</span>.config.source_table_names {
|
|
<a href=#595 id=595 data-nosnippet>595</a> <span class="kw">let </span>table = <span class="self">self
|
|
<a href=#596 id=596 data-nosnippet>596</a> </span>.config
|
|
<a href=#597 id=597 data-nosnippet>597</a> .catalog_manager
|
|
<a href=#598 id=598 data-nosnippet>598</a> .table(<span class="kw-2">&</span>name[<span class="number">0</span>], <span class="kw-2">&</span>name[<span class="number">1</span>], <span class="kw-2">&</span>name[<span class="number">2</span>], <span class="prelude-val">None</span>)
|
|
<a href=#599 id=599 data-nosnippet>599</a> .<span class="kw">await
|
|
<a href=#600 id=600 data-nosnippet>600</a> </span>.map_err(BoxedError::new)
|
|
<a href=#601 id=601 data-nosnippet>601</a> .context(ExternalSnafu)<span class="question-mark">?
|
|
<a href=#602 id=602 data-nosnippet>602</a> </span>.context(UnexpectedSnafu {
|
|
<a href=#603 id=603 data-nosnippet>603</a> reason: <span class="macro">format!</span>(
|
|
<a href=#604 id=604 data-nosnippet>604</a> <span class="string">"Flow {} source table {} is unavailable for recovery retention validation"</span>,
|
|
<a href=#605 id=605 data-nosnippet>605</a> <span class="self">self</span>.config.flow_id,
|
|
<a href=#606 id=606 data-nosnippet>606</a> name.join(<span class="string">"."</span>)
|
|
<a href=#607 id=607 data-nosnippet>607</a> ),
|
|
<a href=#608 id=608 data-nosnippet>608</a> })<span class="question-mark">?</span>;
|
|
<a href=#609 id=609 data-nosnippet>609</a> <span class="kw">let </span>ttl = <span class="kw">if let </span><span class="prelude-val">Some</span>(ttl) = table.table_info().meta.options.ttl {
|
|
<a href=#610 id=610 data-nosnippet>610</a> ttl
|
|
<a href=#611 id=611 data-nosnippet>611</a> } <span class="kw">else </span>{
|
|
<a href=#612 id=612 data-nosnippet>612</a> <span class="kw">let </span>manager = <span class="self">self
|
|
<a href=#613 id=613 data-nosnippet>613</a> </span>.config
|
|
<a href=#614 id=614 data-nosnippet>614</a> .catalog_manager
|
|
<a href=#615 id=615 data-nosnippet>615</a> .as_any()
|
|
<a href=#616 id=616 data-nosnippet>616</a> .downcast_ref::<KvBackendCatalogManager>()
|
|
<a href=#617 id=617 data-nosnippet>617</a> .context(UnexpectedSnafu {
|
|
<a href=#618 id=618 data-nosnippet>618</a> reason: <span class="macro">format!</span>(
|
|
<a href=#619 id=619 data-nosnippet>619</a> <span class="string">"Flow {} cannot resolve inherited TTL for source table {} during recovery"</span>,
|
|
<a href=#620 id=620 data-nosnippet>620</a> <span class="self">self</span>.config.flow_id,
|
|
<a href=#621 id=621 data-nosnippet>621</a> name.join(<span class="string">"."</span>)
|
|
<a href=#622 id=622 data-nosnippet>622</a> ),
|
|
<a href=#623 id=623 data-nosnippet>623</a> })<span class="question-mark">?</span>;
|
|
<a href=#624 id=624 data-nosnippet>624</a> manager
|
|
<a href=#625 id=625 data-nosnippet>625</a> .table_metadata_manager_ref()
|
|
<a href=#626 id=626 data-nosnippet>626</a> .schema_manager()
|
|
<a href=#627 id=627 data-nosnippet>627</a> .get(SchemaNameKey::new(<span class="kw-2">&</span>name[<span class="number">0</span>], <span class="kw-2">&</span>name[<span class="number">1</span>]))
|
|
<a href=#628 id=628 data-nosnippet>628</a> .<span class="kw">await
|
|
<a href=#629 id=629 data-nosnippet>629</a> </span>.map_err(BoxedError::new)
|
|
<a href=#630 id=630 data-nosnippet>630</a> .context(ExternalSnafu)<span class="question-mark">?
|
|
<a href=#631 id=631 data-nosnippet>631</a> </span>.context(UnexpectedSnafu {
|
|
<a href=#632 id=632 data-nosnippet>632</a> reason: <span class="macro">format!</span>(
|
|
<a href=#633 id=633 data-nosnippet>633</a> <span class="string">"Flow {} schema {}.{} is unavailable for recovery retention validation"</span>,
|
|
<a href=#634 id=634 data-nosnippet>634</a> <span class="self">self</span>.config.flow_id, name[<span class="number">0</span>], name[<span class="number">1</span>]
|
|
<a href=#635 id=635 data-nosnippet>635</a> ),
|
|
<a href=#636 id=636 data-nosnippet>636</a> })<span class="question-mark">?
|
|
<a href=#637 id=637 data-nosnippet>637</a> </span>.ttl
|
|
<a href=#638 id=638 data-nosnippet>638</a> .map(Into::into)
|
|
<a href=#639 id=639 data-nosnippet>639</a> .unwrap_or(TimeToLive::Forever)
|
|
<a href=#640 id=640 data-nosnippet>640</a> };
|
|
<a href=#641 id=641 data-nosnippet>641</a>
|
|
<a href=#642 id=642 data-nosnippet>642</a> <span class="kw">match </span>ttl {
|
|
<a href=#643 id=643 data-nosnippet>643</a> TimeToLive::Forever => {}
|
|
<a href=#644 id=644 data-nosnippet>644</a> TimeToLive::Instant => {
|
|
<a href=#645 id=645 data-nosnippet>645</a> <span class="kw">return </span>UnexpectedSnafu {
|
|
<a href=#646 id=646 data-nosnippet>646</a> reason: <span class="macro">format!</span>(
|
|
<a href=#647 id=647 data-nosnippet>647</a> <span class="string">"Flow {} source table {} has instant TTL and cannot be recovered"</span>,
|
|
<a href=#648 id=648 data-nosnippet>648</a> <span class="self">self</span>.config.flow_id,
|
|
<a href=#649 id=649 data-nosnippet>649</a> name.join(<span class="string">"."</span>)
|
|
<a href=#650 id=650 data-nosnippet>650</a> ),
|
|
<a href=#651 id=651 data-nosnippet>651</a> }
|
|
<a href=#652 id=652 data-nosnippet>652</a> .fail();
|
|
<a href=#653 id=653 data-nosnippet>653</a> }
|
|
<a href=#654 id=654 data-nosnippet>654</a> TimeToLive::Duration(ttl) => {
|
|
<a href=#655 id=655 data-nosnippet>655</a> <span class="kw">let </span>expire_after = <span class="self">self</span>.config.expire_after.context(UnexpectedSnafu {
|
|
<a href=#656 id=656 data-nosnippet>656</a> reason: <span class="macro">format!</span>(
|
|
<a href=#657 id=657 data-nosnippet>657</a> <span class="string">"Flow {} source table {} has TTL {ttl:?}, but recovery has no expire_after"</span>,
|
|
<a href=#658 id=658 data-nosnippet>658</a> <span class="self">self</span>.config.flow_id,
|
|
<a href=#659 id=659 data-nosnippet>659</a> name.join(<span class="string">"."</span>)
|
|
<a href=#660 id=660 data-nosnippet>660</a> ),
|
|
<a href=#661 id=661 data-nosnippet>661</a> })<span class="question-mark">?</span>;
|
|
<a href=#662 id=662 data-nosnippet>662</a> <span class="kw">let </span>expire_after = u64::try_from(expire_after).map_err(|<span class="kw">_</span>| {
|
|
<a href=#663 id=663 data-nosnippet>663</a> UnexpectedSnafu {
|
|
<a href=#664 id=664 data-nosnippet>664</a> reason: <span class="macro">format!</span>(
|
|
<a href=#665 id=665 data-nosnippet>665</a> <span class="string">"Flow {} has negative expire_after {expire_after} during recovery retention validation"</span>,
|
|
<a href=#666 id=666 data-nosnippet>666</a> <span class="self">self</span>.config.flow_id
|
|
<a href=#667 id=667 data-nosnippet>667</a> ),
|
|
<a href=#668 id=668 data-nosnippet>668</a> }
|
|
<a href=#669 id=669 data-nosnippet>669</a> .build()
|
|
<a href=#670 id=670 data-nosnippet>670</a> })<span class="question-mark">?</span>;
|
|
<a href=#671 id=671 data-nosnippet>671</a> <span class="kw">if </span>ttl <= Duration::from_secs(expire_after) {
|
|
<a href=#672 id=672 data-nosnippet>672</a> <span class="kw">return </span>UnexpectedSnafu {
|
|
<a href=#673 id=673 data-nosnippet>673</a> reason: <span class="macro">format!</span>(
|
|
<a href=#674 id=674 data-nosnippet>674</a> <span class="string">"Flow {} source table {} TTL {ttl:?} must exceed expire_after {expire_after}s for recovery"</span>,
|
|
<a href=#675 id=675 data-nosnippet>675</a> <span class="self">self</span>.config.flow_id,
|
|
<a href=#676 id=676 data-nosnippet>676</a> name.join(<span class="string">"."</span>)
|
|
<a href=#677 id=677 data-nosnippet>677</a> ),
|
|
<a href=#678 id=678 data-nosnippet>678</a> }
|
|
<a href=#679 id=679 data-nosnippet>679</a> .fail();
|
|
<a href=#680 id=680 data-nosnippet>680</a> }
|
|
<a href=#681 id=681 data-nosnippet>681</a> <span class="kw">let </span><span class="kw-2">mut </span>oldest = retention_lower.context(UnexpectedSnafu {
|
|
<a href=#682 id=682 data-nosnippet>682</a> reason: <span class="macro">format!</span>(
|
|
<a href=#683 id=683 data-nosnippet>683</a> <span class="string">"Flow {} source table {} has finite TTL {ttl:?}, but recovery has no retained lower bound"</span>,
|
|
<a href=#684 id=684 data-nosnippet>684</a> <span class="self">self</span>.config.flow_id,
|
|
<a href=#685 id=685 data-nosnippet>685</a> name.join(<span class="string">"."</span>)
|
|
<a href=#686 id=686 data-nosnippet>686</a> ),
|
|
<a href=#687 id=687 data-nosnippet>687</a> })<span class="question-mark">?</span>;
|
|
<a href=#688 id=688 data-nosnippet>688</a> <span class="kw">for </span>(start, <span class="kw">_</span>) <span class="kw">in </span>windows {
|
|
<a href=#689 id=689 data-nosnippet>689</a> oldest = oldest.min(<span class="kw-2">*</span>start);
|
|
<a href=#690 id=690 data-nosnippet>690</a> }
|
|
<a href=#691 id=691 data-nosnippet>691</a> <span class="kw">let </span>cutoff = Timestamp::current_millis()
|
|
<a href=#692 id=692 data-nosnippet>692</a> .sub_duration(ttl)
|
|
<a href=#693 id=693 data-nosnippet>693</a> .map_err(BoxedError::new)
|
|
<a href=#694 id=694 data-nosnippet>694</a> .context(ExternalSnafu)<span class="question-mark">?</span>;
|
|
<a href=#695 id=695 data-nosnippet>695</a> <span class="kw">if </span>oldest <= cutoff {
|
|
<a href=#696 id=696 data-nosnippet>696</a> <span class="kw">return </span>UnexpectedSnafu {
|
|
<a href=#697 id=697 data-nosnippet>697</a> reason: <span class="macro">format!</span>(
|
|
<a href=#698 id=698 data-nosnippet>698</a> <span class="string">"Flow {} recovery requires source table {} data from {oldest:?}, older than TTL {ttl:?} cutoff {cutoff:?}"</span>,
|
|
<a href=#699 id=699 data-nosnippet>699</a> <span class="self">self</span>.config.flow_id,
|
|
<a href=#700 id=700 data-nosnippet>700</a> name.join(<span class="string">"."</span>)
|
|
<a href=#701 id=701 data-nosnippet>701</a> ),
|
|
<a href=#702 id=702 data-nosnippet>702</a> }
|
|
<a href=#703 id=703 data-nosnippet>703</a> .fail();
|
|
<a href=#704 id=704 data-nosnippet>704</a> }
|
|
<a href=#705 id=705 data-nosnippet>705</a> }
|
|
<a href=#706 id=706 data-nosnippet>706</a> }
|
|
<a href=#707 id=707 data-nosnippet>707</a> }
|
|
<a href=#708 id=708 data-nosnippet>708</a> <span class="prelude-val">Ok</span>(())
|
|
<a href=#709 id=709 data-nosnippet>709</a> }
|
|
<a href=#710 id=710 data-nosnippet>710</a>
|
|
<a href=#711 id=711 data-nosnippet>711</a> <span class="doccomment">/// Discovers the source time windows touched by the exact sequence range `(C, H]`.
|
|
<a href=#712 id=712 data-nosnippet>712</a> ///
|
|
<a href=#713 id=713 data-nosnippet>713</a> /// This compatibility wrapper captures the retention bound before discovery.
|
|
<a href=#714 id=714 data-nosnippet>714</a> </span><span class="kw">pub async fn </span>capture_recovery_windows(
|
|
<a href=#715 id=715 data-nosnippet>715</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#716 id=716 data-nosnippet>716</a> engine: <span class="kw-2">&</span>QueryEngineRef,
|
|
<a href=#717 id=717 data-nosnippet>717</a> frontend_client: <span class="kw-2">&</span>FrontendClient,
|
|
<a href=#718 id=718 data-nosnippet>718</a> lower: <span class="kw-2">&</span>BTreeMap<u64, u64>,
|
|
<a href=#719 id=719 data-nosnippet>719</a> ) -> <span class="prelude-ty">Result</span><(BTreeMap<u64, u64>, Vec<(Timestamp, Timestamp)>), Error> {
|
|
<a href=#720 id=720 data-nosnippet>720</a> <span class="kw">let </span>retention_lower = <span class="self">self</span>.recovery_retention_lower_bound()<span class="question-mark">?</span>;
|
|
<a href=#721 id=721 data-nosnippet>721</a> <span class="self">self</span>.capture_recovery_windows_since(engine, frontend_client, lower, retention_lower)
|
|
<a href=#722 id=722 data-nosnippet>722</a> .<span class="kw">await
|
|
<a href=#723 id=723 data-nosnippet>723</a> </span>}
|
|
<a href=#724 id=724 data-nosnippet>724</a>
|
|
<a href=#725 id=725 data-nosnippet>725</a> <span class="doccomment">/// Discovers windows under `lower` using a caller-frozen retention bound.
|
|
<a href=#726 id=726 data-nosnippet>726</a> ///
|
|
<a href=#727 id=727 data-nosnippet>727</a> /// The caller owns the execution guard and freezes this bound with its recovery scope.
|
|
<a href=#728 id=728 data-nosnippet>728</a> /// This read-only method leaves task state untouched. Terminal proof must cover every
|
|
<a href=#729 id=729 data-nosnippet>729</a> /// region in `lower`; subset or pruned proofs are rejected.
|
|
<a href=#730 id=730 data-nosnippet>730</a> </span><span class="kw">pub async fn </span>capture_recovery_windows_since(
|
|
<a href=#731 id=731 data-nosnippet>731</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#732 id=732 data-nosnippet>732</a> engine: <span class="kw-2">&</span>QueryEngineRef,
|
|
<a href=#733 id=733 data-nosnippet>733</a> frontend_client: <span class="kw-2">&</span>FrontendClient,
|
|
<a href=#734 id=734 data-nosnippet>734</a> lower: <span class="kw-2">&</span>BTreeMap<u64, u64>,
|
|
<a href=#735 id=735 data-nosnippet>735</a> retention_lower: <span class="prelude-ty">Option</span><Timestamp>,
|
|
<a href=#736 id=736 data-nosnippet>736</a> ) -> <span class="prelude-ty">Result</span><(BTreeMap<u64, u64>, Vec<(Timestamp, Timestamp)>), Error> {
|
|
<a href=#737 id=737 data-nosnippet>737</a> <span class="kw">if </span>lower.is_empty() {
|
|
<a href=#738 id=738 data-nosnippet>738</a> <span class="kw">return </span>UnexpectedSnafu {
|
|
<a href=#739 id=739 data-nosnippet>739</a> reason: <span class="macro">format!</span>(
|
|
<a href=#740 id=740 data-nosnippet>740</a> <span class="string">"Flow {} recovery window capture requires nonempty lower sequence bounds"</span>,
|
|
<a href=#741 id=741 data-nosnippet>741</a> <span class="self">self</span>.config.flow_id
|
|
<a href=#742 id=742 data-nosnippet>742</a> ),
|
|
<a href=#743 id=743 data-nosnippet>743</a> }
|
|
<a href=#744 id=744 data-nosnippet>744</a> .fail();
|
|
<a href=#745 id=745 data-nosnippet>745</a> }
|
|
<a href=#746 id=746 data-nosnippet>746</a> <span class="kw">if </span>!<span class="self">self</span>.sequence_range_capable().<span class="kw">await</span><span class="question-mark">? </span>{
|
|
<a href=#747 id=747 data-nosnippet>747</a> <span class="kw">return </span>UnexpectedSnafu {
|
|
<a href=#748 id=748 data-nosnippet>748</a> reason: <span class="macro">format!</span>(
|
|
<a href=#749 id=749 data-nosnippet>749</a> <span class="string">"Flow {} recovery window capture requires sequence_range-capable sources"</span>,
|
|
<a href=#750 id=750 data-nosnippet>750</a> <span class="self">self</span>.config.flow_id
|
|
<a href=#751 id=751 data-nosnippet>751</a> ),
|
|
<a href=#752 id=752 data-nosnippet>752</a> }
|
|
<a href=#753 id=753 data-nosnippet>753</a> .fail();
|
|
<a href=#754 id=754 data-nosnippet>754</a> }
|
|
<a href=#755 id=755 data-nosnippet>755</a>
|
|
<a href=#756 id=756 data-nosnippet>756</a> <span class="kw">let </span>query_ctx = <span class="self">self</span>.query_context_snapshot();
|
|
<a href=#757 id=757 data-nosnippet>757</a> <span class="kw">let </span>plan = sql_to_df_plan(query_ctx, engine.clone(), <span class="kw-2">&</span><span class="self">self</span>.config.query, <span class="bool-val">false</span>).<span class="kw">await</span><span class="question-mark">?</span>;
|
|
<a href=#758 id=758 data-nosnippet>758</a> <span class="kw">let </span>input = recovery_aggregate_input(<span class="kw-2">&</span>plan)<span class="question-mark">?</span>;
|
|
<a href=#759 id=759 data-nosnippet>759</a> <span class="kw">let </span><span class="prelude-val">Some</span>(analysis) = analyze_incremental_aggregate_plan(<span class="kw-2">&</span>plan)<span class="question-mark">? </span><span class="kw">else </span>{
|
|
<a href=#760 id=760 data-nosnippet>760</a> <span class="kw">return </span>UnexpectedSnafu {
|
|
<a href=#761 id=761 data-nosnippet>761</a> reason: <span class="macro">format!</span>(
|
|
<a href=#762 id=762 data-nosnippet>762</a> <span class="string">"Flow {} recovery window capture requires a supported aggregate input"</span>,
|
|
<a href=#763 id=763 data-nosnippet>763</a> <span class="self">self</span>.config.flow_id
|
|
<a href=#764 id=764 data-nosnippet>764</a> ),
|
|
<a href=#765 id=765 data-nosnippet>765</a> }
|
|
<a href=#766 id=766 data-nosnippet>766</a> .fail();
|
|
<a href=#767 id=767 data-nosnippet>767</a> };
|
|
<a href=#768 id=768 data-nosnippet>768</a> <span class="kw">if </span>!analysis.unsupported_exprs.is_empty() {
|
|
<a href=#769 id=769 data-nosnippet>769</a> <span class="kw">return </span>UnexpectedSnafu {
|
|
<a href=#770 id=770 data-nosnippet>770</a> reason: <span class="macro">format!</span>(
|
|
<a href=#771 id=771 data-nosnippet>771</a> <span class="string">"Flow {} recovery window capture has unsupported aggregate expressions: {:?}"</span>,
|
|
<a href=#772 id=772 data-nosnippet>772</a> <span class="self">self</span>.config.flow_id, analysis.unsupported_exprs
|
|
<a href=#773 id=773 data-nosnippet>773</a> ),
|
|
<a href=#774 id=774 data-nosnippet>774</a> }
|
|
<a href=#775 id=775 data-nosnippet>775</a> .fail();
|
|
<a href=#776 id=776 data-nosnippet>776</a> }
|
|
<a href=#777 id=777 data-nosnippet>777</a> <span class="kw">let </span>time_window_expr = <span class="self">self
|
|
<a href=#778 id=778 data-nosnippet>778</a> </span>.config
|
|
<a href=#779 id=779 data-nosnippet>779</a> .time_window_expr
|
|
<a href=#780 id=780 data-nosnippet>780</a> .as_ref()
|
|
<a href=#781 id=781 data-nosnippet>781</a> .context(UnexpectedSnafu {
|
|
<a href=#782 id=782 data-nosnippet>782</a> reason: <span class="macro">format!</span>(
|
|
<a href=#783 id=783 data-nosnippet>783</a> <span class="string">"Flow {} recovery window capture requires a time-window expression"</span>,
|
|
<a href=#784 id=784 data-nosnippet>784</a> <span class="self">self</span>.config.flow_id
|
|
<a href=#785 id=785 data-nosnippet>785</a> ),
|
|
<a href=#786 id=786 data-nosnippet>786</a> })<span class="question-mark">?</span>;
|
|
<a href=#787 id=787 data-nosnippet>787</a> <span class="kw">let </span>input = <span class="kw">if let </span><span class="prelude-val">Some</span>(retention_lower) = retention_lower {
|
|
<a href=#788 id=788 data-nosnippet>788</a> <span class="kw">let </span><span class="kw-2">mut </span>add_filter = AddFilterRewriter::new(
|
|
<a href=#789 id=789 data-nosnippet>789</a> col(<span class="kw-2">&</span>time_window_expr.column_name).gt_eq(lit(to_df_literal(retention_lower)<span class="question-mark">?</span>)),
|
|
<a href=#790 id=790 data-nosnippet>790</a> );
|
|
<a href=#791 id=791 data-nosnippet>791</a> input
|
|
<a href=#792 id=792 data-nosnippet>792</a> .rewrite(<span class="kw-2">&mut </span>add_filter)
|
|
<a href=#793 id=793 data-nosnippet>793</a> .with_context(|<span class="kw">_</span>| DatafusionSnafu {
|
|
<a href=#794 id=794 data-nosnippet>794</a> context: <span class="string">"Failed to apply recovery expire_after filter"</span>.to_string(),
|
|
<a href=#795 id=795 data-nosnippet>795</a> })<span class="question-mark">?
|
|
<a href=#796 id=796 data-nosnippet>796</a> </span>.data
|
|
<a href=#797 id=797 data-nosnippet>797</a> } <span class="kw">else </span>{
|
|
<a href=#798 id=798 data-nosnippet>798</a> input
|
|
<a href=#799 id=799 data-nosnippet>799</a> };
|
|
<a href=#800 id=800 data-nosnippet>800</a> <span class="kw">let </span>timestamp_plan = LogicalPlan::Projection(
|
|
<a href=#801 id=801 data-nosnippet>801</a> Projection::try_new(<span class="macro">vec!</span>[col(<span class="kw-2">&</span>time_window_expr.column_name)], Arc::new(input))
|
|
<a href=#802 id=802 data-nosnippet>802</a> .context(DatafusionSnafu {
|
|
<a href=#803 id=803 data-nosnippet>803</a> context: <span class="string">"Failed to project recovery source timestamps"</span>.to_string(),
|
|
<a href=#804 id=804 data-nosnippet>804</a> })<span class="question-mark">?</span>,
|
|
<a href=#805 id=805 data-nosnippet>805</a> );
|
|
<a href=#806 id=806 data-nosnippet>806</a> <span class="kw">let </span>catalog = <span class="kw-2">&</span><span class="self">self</span>.config.sink_table_name[<span class="number">0</span>];
|
|
<a href=#807 id=807 data-nosnippet>807</a> <span class="kw">let </span>schema = <span class="kw-2">&</span><span class="self">self</span>.config.sink_table_name[<span class="number">1</span>];
|
|
<a href=#808 id=808 data-nosnippet>808</a> <span class="kw">let </span>timestamp_plan = timestamp_plan
|
|
<a href=#809 id=809 data-nosnippet>809</a> .clone()
|
|
<a href=#810 id=810 data-nosnippet>810</a> .transform_down_with_subqueries(|p| {
|
|
<a href=#811 id=811 data-nosnippet>811</a> <span class="kw">if let </span>LogicalPlan::TableScan(<span class="kw-2">mut </span>table_scan) = p {
|
|
<a href=#812 id=812 data-nosnippet>812</a> <span class="kw">let </span>resolved = table_scan.table_name.resolve(catalog, schema);
|
|
<a href=#813 id=813 data-nosnippet>813</a> table_scan.table_name = resolved.into();
|
|
<a href=#814 id=814 data-nosnippet>814</a> <span class="prelude-val">Ok</span>(Transformed::yes(LogicalPlan::TableScan(table_scan)))
|
|
<a href=#815 id=815 data-nosnippet>815</a> } <span class="kw">else </span>{
|
|
<a href=#816 id=816 data-nosnippet>816</a> <span class="prelude-val">Ok</span>(Transformed::no(p))
|
|
<a href=#817 id=817 data-nosnippet>817</a> }
|
|
<a href=#818 id=818 data-nosnippet>818</a> })
|
|
<a href=#819 id=819 data-nosnippet>819</a> .with_context(|<span class="kw">_</span>| DatafusionSnafu {
|
|
<a href=#820 id=820 data-nosnippet>820</a> context: <span class="macro">format!</span>(
|
|
<a href=#821 id=821 data-nosnippet>821</a> <span class="string">"Failed to fix table ref in recovery timestamp plan, plan={:?}"</span>,
|
|
<a href=#822 id=822 data-nosnippet>822</a> timestamp_plan
|
|
<a href=#823 id=823 data-nosnippet>823</a> ),
|
|
<a href=#824 id=824 data-nosnippet>824</a> })<span class="question-mark">?
|
|
<a href=#825 id=825 data-nosnippet>825</a> </span>.data;
|
|
<a href=#826 id=826 data-nosnippet>826</a> <span class="kw">let </span>message = DFLogicalSubstraitConvertor {}
|
|
<a href=#827 id=827 data-nosnippet>827</a> .encode(<span class="kw-2">&</span>timestamp_plan, DefaultSerializer)
|
|
<a href=#828 id=828 data-nosnippet>828</a> .context(SubstraitEncodeLogicalPlanSnafu)<span class="question-mark">?</span>;
|
|
<a href=#829 id=829 data-nosnippet>829</a> <span class="kw">let </span>request = api::v1::QueryRequest {
|
|
<a href=#830 id=830 data-nosnippet>830</a> query: <span class="prelude-val">Some</span>(api::v1::query_request::Query::LogicalPlan(message.to_vec())),
|
|
<a href=#831 id=831 data-nosnippet>831</a> };
|
|
<a href=#832 id=832 data-nosnippet>832</a> <span class="kw">let </span>lower_json = serde_json::to_string(lower).map_err(|err| {
|
|
<a href=#833 id=833 data-nosnippet>833</a> UnexpectedSnafu {
|
|
<a href=#834 id=834 data-nosnippet>834</a> reason: <span class="macro">format!</span>(<span class="string">"Failed to serialize recovery lower sequence bounds: {err}"</span>),
|
|
<a href=#835 id=835 data-nosnippet>835</a> }
|
|
<a href=#836 id=836 data-nosnippet>836</a> .build()
|
|
<a href=#837 id=837 data-nosnippet>837</a> })<span class="question-mark">?</span>;
|
|
<a href=#838 id=838 data-nosnippet>838</a> <span class="kw">let </span>extensions = [
|
|
<a href=#839 id=839 data-nosnippet>839</a> (FLOW_RETURN_REGION_SEQ, <span class="string">"true"</span>),
|
|
<a href=#840 id=840 data-nosnippet>840</a> (FLOW_INCREMENTAL_MODE, FLOW_INCREMENTAL_MODE_SEQUENCE_RANGE),
|
|
<a href=#841 id=841 data-nosnippet>841</a> (FLOW_INCREMENTAL_AFTER_SEQS, lower_json.as_str()),
|
|
<a href=#842 id=842 data-nosnippet>842</a> ];
|
|
<a href=#843 id=843 data-nosnippet>843</a> <span class="kw">let </span><span class="kw-2">mut </span>peer_desc = <span class="prelude-val">None</span>;
|
|
<a href=#844 id=844 data-nosnippet>844</a> <span class="kw">let </span>result = frontend_client
|
|
<a href=#845 id=845 data-nosnippet>845</a> .query_with_terminal_metrics(
|
|
<a href=#846 id=846 data-nosnippet>846</a> catalog,
|
|
<a href=#847 id=847 data-nosnippet>847</a> schema,
|
|
<a href=#848 id=848 data-nosnippet>848</a> request,
|
|
<a href=#849 id=849 data-nosnippet>849</a> <span class="kw-2">&</span>extensions,
|
|
<a href=#850 id=850 data-nosnippet>850</a> <span class="kw-2">&</span>HashMap::new(),
|
|
<a href=#851 id=851 data-nosnippet>851</a> <span class="kw-2">&mut </span>peer_desc,
|
|
<a href=#852 id=852 data-nosnippet>852</a> )
|
|
<a href=#853 id=853 data-nosnippet>853</a> .<span class="kw">await</span><span class="question-mark">?</span>;
|
|
<a href=#854 id=854 data-nosnippet>854</a> <span class="kw">let </span><span class="kw-2">mut </span>aligned_windows = BTreeSet::new();
|
|
<a href=#855 id=855 data-nosnippet>855</a> <span class="kw">let </span>metrics = result.metrics.clone();
|
|
<a href=#856 id=856 data-nosnippet>856</a> <span class="kw">match </span>result.output.data {
|
|
<a href=#857 id=857 data-nosnippet>857</a> OutputData::AffectedRows(<span class="kw">_</span>) => {
|
|
<a href=#858 id=858 data-nosnippet>858</a> <span class="kw">return </span>UnexpectedSnafu {
|
|
<a href=#859 id=859 data-nosnippet>859</a> reason: <span class="string">"Recovery timestamp projection unexpectedly returned affected rows"
|
|
<a href=#860 id=860 data-nosnippet>860</a> </span>.to_string(),
|
|
<a href=#861 id=861 data-nosnippet>861</a> }
|
|
<a href=#862 id=862 data-nosnippet>862</a> .fail();
|
|
<a href=#863 id=863 data-nosnippet>863</a> }
|
|
<a href=#864 id=864 data-nosnippet>864</a> OutputData::RecordBatches(batches) => {
|
|
<a href=#865 id=865 data-nosnippet>865</a> <span class="kw">for </span>batch <span class="kw">in </span>batches.iter() {
|
|
<a href=#866 id=866 data-nosnippet>866</a> capture_recovery_batch_windows(batch, time_window_expr, <span class="kw-2">&mut </span>aligned_windows)<span class="question-mark">?</span>;
|
|
<a href=#867 id=867 data-nosnippet>867</a> }
|
|
<a href=#868 id=868 data-nosnippet>868</a> }
|
|
<a href=#869 id=869 data-nosnippet>869</a> OutputData::Stream(<span class="kw-2">mut </span>stream) => {
|
|
<a href=#870 id=870 data-nosnippet>870</a> <span class="kw">while let </span><span class="prelude-val">Some</span>(batch) = stream
|
|
<a href=#871 id=871 data-nosnippet>871</a> .try_next()
|
|
<a href=#872 id=872 data-nosnippet>872</a> .<span class="kw">await
|
|
<a href=#873 id=873 data-nosnippet>873</a> </span>.map_err(BoxedError::new)
|
|
<a href=#874 id=874 data-nosnippet>874</a> .context(ExternalSnafu)<span class="question-mark">?
|
|
<a href=#875 id=875 data-nosnippet>875</a> </span>{
|
|
<a href=#876 id=876 data-nosnippet>876</a> capture_recovery_batch_windows(<span class="kw-2">&</span>batch, time_window_expr, <span class="kw-2">&mut </span>aligned_windows)<span class="question-mark">?</span>;
|
|
<a href=#877 id=877 data-nosnippet>877</a> }
|
|
<a href=#878 id=878 data-nosnippet>878</a> }
|
|
<a href=#879 id=879 data-nosnippet>879</a> }
|
|
<a href=#880 id=880 data-nosnippet>880</a> <span class="kw">if </span>!metrics.is_ready() {
|
|
<a href=#881 id=881 data-nosnippet>881</a> <span class="kw">return </span>UnexpectedSnafu {
|
|
<a href=#882 id=882 data-nosnippet>882</a> reason: <span class="string">"Recovery timestamp projection ended without terminal metrics"</span>.to_string(),
|
|
<a href=#883 id=883 data-nosnippet>883</a> }
|
|
<a href=#884 id=884 data-nosnippet>884</a> .fail();
|
|
<a href=#885 id=885 data-nosnippet>885</a> }
|
|
<a href=#886 id=886 data-nosnippet>886</a> <span class="kw">let </span>participating = metrics.participating_regions().context(UnexpectedSnafu {
|
|
<a href=#887 id=887 data-nosnippet>887</a> reason: <span class="string">"Recovery timestamp projection has no participating-region proof"</span>.to_string(),
|
|
<a href=#888 id=888 data-nosnippet>888</a> })<span class="question-mark">?</span>;
|
|
<a href=#889 id=889 data-nosnippet>889</a> <span class="kw">let </span>high = metrics.region_watermark_map().context(UnexpectedSnafu {
|
|
<a href=#890 id=890 data-nosnippet>890</a> reason: <span class="string">"Recovery timestamp projection has no terminal watermark proof"</span>.to_string(),
|
|
<a href=#891 id=891 data-nosnippet>891</a> })<span class="question-mark">?</span>;
|
|
<a href=#892 id=892 data-nosnippet>892</a> <span class="kw">if </span>participating.len() != lower.len()
|
|
<a href=#893 id=893 data-nosnippet>893</a> || high.len() != lower.len()
|
|
<a href=#894 id=894 data-nosnippet>894</a> || participating
|
|
<a href=#895 id=895 data-nosnippet>895</a> .iter()
|
|
<a href=#896 id=896 data-nosnippet>896</a> .any(|region| !lower.contains_key(region))
|
|
<a href=#897 id=897 data-nosnippet>897</a> || lower
|
|
<a href=#898 id=898 data-nosnippet>898</a> .iter()
|
|
<a href=#899 id=899 data-nosnippet>899</a> .any(|(region, low)| high.get(region).is_none_or(|watermark| watermark < low))
|
|
<a href=#900 id=900 data-nosnippet>900</a> || high.keys().any(|region| !participating.contains(region))
|
|
<a href=#901 id=901 data-nosnippet>901</a> {
|
|
<a href=#902 id=902 data-nosnippet>902</a> <span class="kw">return </span>UnexpectedSnafu {
|
|
<a href=#903 id=903 data-nosnippet>903</a> reason: <span class="macro">format!</span>(
|
|
<a href=#904 id=904 data-nosnippet>904</a> <span class="string">"Recovery timestamp projection returned incomplete or regressing terminal proof: lower={lower:?}, participating={participating:?}, high={high:?}"
|
|
<a href=#905 id=905 data-nosnippet>905</a> </span>),
|
|
<a href=#906 id=906 data-nosnippet>906</a> }
|
|
<a href=#907 id=907 data-nosnippet>907</a> .fail();
|
|
<a href=#908 id=908 data-nosnippet>908</a> }
|
|
<a href=#909 id=909 data-nosnippet>909</a> <span class="prelude-val">Ok</span>((
|
|
<a href=#910 id=910 data-nosnippet>910</a> high.into_iter().collect(),
|
|
<a href=#911 id=911 data-nosnippet>911</a> aligned_windows.into_iter().collect(),
|
|
<a href=#912 id=912 data-nosnippet>912</a> ))
|
|
<a href=#913 id=913 data-nosnippet>913</a> }
|
|
<a href=#914 id=914 data-nosnippet>914</a>
|
|
<a href=#915 id=915 data-nosnippet>915</a> <span class="doccomment">/// Validates that the sink table schema can accept this flow's ordinary output.
|
|
<a href=#916 id=916 data-nosnippet>916</a> </span><span class="kw">pub async fn </span>validate_sink_table_schema(<span class="kw-2">&</span><span class="self">self</span>, engine: <span class="kw-2">&</span>QueryEngineRef) -> <span class="prelude-ty">Result</span><(), Error> {
|
|
<a href=#917 id=917 data-nosnippet>917</a> <span class="self">self</span>.validate_sink_table_schema_with_values(engine, <span class="kw-2">&</span>BTreeMap::new())
|
|
<a href=#918 id=918 data-nosnippet>918</a> .<span class="kw">await
|
|
<a href=#919 id=919 data-nosnippet>919</a> </span>}
|
|
<a href=#920 id=920 data-nosnippet>920</a>
|
|
<a href=#921 id=921 data-nosnippet>921</a> <span class="doccomment">/// Validates the flow output plus extension-owned typed sink values without consuming dirty work.
|
|
<a href=#922 id=922 data-nosnippet>922</a> </span><span class="kw">pub async fn </span>validate_sink_table_schema_with_values(
|
|
<a href=#923 id=923 data-nosnippet>923</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#924 id=924 data-nosnippet>924</a> engine: <span class="kw-2">&</span>QueryEngineRef,
|
|
<a href=#925 id=925 data-nosnippet>925</a> values: <span class="kw-2">&</span>BTreeMap<String, ScalarValue>,
|
|
<a href=#926 id=926 data-nosnippet>926</a> ) -> <span class="prelude-ty">Result</span><(), Error> {
|
|
<a href=#927 id=927 data-nosnippet>927</a> <span class="kw">let </span>(table, <span class="kw">_</span>) = get_table_info_df_schema(
|
|
<a href=#928 id=928 data-nosnippet>928</a> <span class="self">self</span>.config.catalog_manager.clone(),
|
|
<a href=#929 id=929 data-nosnippet>929</a> <span class="self">self</span>.config.sink_table_name.clone(),
|
|
<a href=#930 id=930 data-nosnippet>930</a> )
|
|
<a href=#931 id=931 data-nosnippet>931</a> .<span class="kw">await</span><span class="question-mark">?</span>;
|
|
<a href=#932 id=932 data-nosnippet>932</a>
|
|
<a href=#933 id=933 data-nosnippet>933</a> <span class="kw">let </span>table_meta = <span class="kw-2">&</span>table.table_info().meta;
|
|
<a href=#934 id=934 data-nosnippet>934</a> <span class="kw">let </span>merge_mode_last_non_null =
|
|
<a href=#935 id=935 data-nosnippet>935</a> is_merge_mode_last_non_null(<span class="kw-2">&</span>table_meta.options.extra_options);
|
|
<a href=#936 id=936 data-nosnippet>936</a> <span class="kw">let </span>primary_key_indices = table_meta.primary_key_indices.clone();
|
|
<a href=#937 id=937 data-nosnippet>937</a>
|
|
<a href=#938 id=938 data-nosnippet>938</a> gen_plan_with_matching_schema_and_values(
|
|
<a href=#939 id=939 data-nosnippet>939</a> <span class="kw-2">&</span><span class="self">self</span>.config.query,
|
|
<a href=#940 id=940 data-nosnippet>940</a> <span class="self">self</span>.query_context_snapshot(),
|
|
<a href=#941 id=941 data-nosnippet>941</a> engine.clone(),
|
|
<a href=#942 id=942 data-nosnippet>942</a> table_meta.schema.clone(),
|
|
<a href=#943 id=943 data-nosnippet>943</a> <span class="kw-2">&</span>primary_key_indices,
|
|
<a href=#944 id=944 data-nosnippet>944</a> merge_mode_last_non_null,
|
|
<a href=#945 id=945 data-nosnippet>945</a> values,
|
|
<a href=#946 id=946 data-nosnippet>946</a> )
|
|
<a href=#947 id=947 data-nosnippet>947</a> .<span class="kw">await
|
|
<a href=#948 id=948 data-nosnippet>948</a> </span>.map(|<span class="kw">_</span>| ())
|
|
<a href=#949 id=949 data-nosnippet>949</a> }
|
|
<a href=#950 id=950 data-nosnippet>950</a>
|
|
<a href=#951 id=951 data-nosnippet>951</a> <span class="kw">async fn </span>is_table_exist(<span class="kw-2">&</span><span class="self">self</span>, table_name: <span class="kw-2">&</span>[String; <span class="number">3</span>]) -> <span class="prelude-ty">Result</span><bool, Error> {
|
|
<a href=#952 id=952 data-nosnippet>952</a> <span class="self">self</span>.config
|
|
<a href=#953 id=953 data-nosnippet>953</a> .catalog_manager
|
|
<a href=#954 id=954 data-nosnippet>954</a> .table_exists(<span class="kw-2">&</span>table_name[<span class="number">0</span>], <span class="kw-2">&</span>table_name[<span class="number">1</span>], <span class="kw-2">&</span>table_name[<span class="number">2</span>], <span class="prelude-val">None</span>)
|
|
<a href=#955 id=955 data-nosnippet>955</a> .<span class="kw">await
|
|
<a href=#956 id=956 data-nosnippet>956</a> </span>.map_err(BoxedError::new)
|
|
<a href=#957 id=957 data-nosnippet>957</a> .context(ExternalSnafu)
|
|
<a href=#958 id=958 data-nosnippet>958</a> }
|
|
<a href=#959 id=959 data-nosnippet>959</a>
|
|
<a href=#960 id=960 data-nosnippet>960</a> <span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">async fn </span>execute_once_serialized(
|
|
<a href=#961 id=961 data-nosnippet>961</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#962 id=962 data-nosnippet>962</a> engine: <span class="kw-2">&</span>QueryEngineRef,
|
|
<a href=#963 id=963 data-nosnippet>963</a> frontend_client: <span class="kw-2">&</span>Arc<FrontendClient>,
|
|
<a href=#964 id=964 data-nosnippet>964</a> max_window_cnt: <span class="prelude-ty">Option</span><usize>,
|
|
<a href=#965 id=965 data-nosnippet>965</a> ) -> <span class="prelude-ty">Result</span><<span class="prelude-ty">Option</span><(usize, Duration)>, Error> {
|
|
<a href=#966 id=966 data-nosnippet>966</a> <span class="kw">let </span>outcome = <span class="self">self
|
|
<a href=#967 id=967 data-nosnippet>967</a> </span>.execute_once_serialized_with_outcome(engine, frontend_client, max_window_cnt)
|
|
<a href=#968 id=968 data-nosnippet>968</a> .<span class="kw">await</span>;
|
|
<a href=#969 id=969 data-nosnippet>969</a> outcome.result
|
|
<a href=#970 id=970 data-nosnippet>970</a> }
|
|
<a href=#971 id=971 data-nosnippet>971</a>
|
|
<a href=#972 id=972 data-nosnippet>972</a> <span class="doccomment">/// Executes one flow evaluation under `execution_lock` and keeps the
|
|
<a href=#973 id=973 data-nosnippet>973</a> /// generated query context for the background loop's error logging/backoff.
|
|
<a href=#974 id=974 data-nosnippet>974</a> </span><span class="kw">async fn </span>execute_once_serialized_with_outcome(
|
|
<a href=#975 id=975 data-nosnippet>975</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#976 id=976 data-nosnippet>976</a> engine: <span class="kw-2">&</span>QueryEngineRef,
|
|
<a href=#977 id=977 data-nosnippet>977</a> frontend_client: <span class="kw-2">&</span>Arc<FrontendClient>,
|
|
<a href=#978 id=978 data-nosnippet>978</a> max_window_cnt: <span class="prelude-ty">Option</span><usize>,
|
|
<a href=#979 id=979 data-nosnippet>979</a> ) -> ExecuteOnceOutcome {
|
|
<a href=#980 id=980 data-nosnippet>980</a> <span class="kw">let </span>guard = BatchingExecutionGuard::new(<span class="self">self</span>.execution_lock.clone().lock_owned().<span class="kw">await</span>);
|
|
<a href=#981 id=981 data-nosnippet>981</a> <span class="self">self</span>.execute_once_with_guard(guard, engine, frontend_client, max_window_cnt)
|
|
<a href=#982 id=982 data-nosnippet>982</a> .<span class="kw">await
|
|
<a href=#983 id=983 data-nosnippet>983</a> </span>}
|
|
<a href=#984 id=984 data-nosnippet>984</a>
|
|
<a href=#985 id=985 data-nosnippet>985</a> <span class="kw">async fn </span>execute_once_with_guard(
|
|
<a href=#986 id=986 data-nosnippet>986</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#987 id=987 data-nosnippet>987</a> guard: BatchingExecutionGuard,
|
|
<a href=#988 id=988 data-nosnippet>988</a> engine: <span class="kw-2">&</span>QueryEngineRef,
|
|
<a href=#989 id=989 data-nosnippet>989</a> frontend_client: <span class="kw-2">&</span>Arc<FrontendClient>,
|
|
<a href=#990 id=990 data-nosnippet>990</a> max_window_cnt: <span class="prelude-ty">Option</span><usize>,
|
|
<a href=#991 id=991 data-nosnippet>991</a> ) -> ExecuteOnceOutcome {
|
|
<a href=#992 id=992 data-nosnippet>992</a> <span class="kw">if let </span><span class="prelude-val">Some</span>(execution) = <span class="kw-2">&</span><span class="self">self</span>.execution {
|
|
<a href=#993 id=993 data-nosnippet>993</a> <span class="kw">return </span>execution
|
|
<a href=#994 id=994 data-nosnippet>994</a> .clone()
|
|
<a href=#995 id=995 data-nosnippet>995</a> .execute_once(guard, <span class="self">self</span>, engine, frontend_client, max_window_cnt)
|
|
<a href=#996 id=996 data-nosnippet>996</a> .<span class="kw">await</span>;
|
|
<a href=#997 id=997 data-nosnippet>997</a> }
|
|
<a href=#998 id=998 data-nosnippet>998</a> <span class="self">self</span>.execute_once_default_unlocked(engine, frontend_client, max_window_cnt)
|
|
<a href=#999 id=999 data-nosnippet>999</a> .<span class="kw">await
|
|
<a href=#1000 id=1000 data-nosnippet>1000</a> </span>}
|
|
<a href=#1001 id=1001 data-nosnippet>1001</a>
|
|
<a href=#1002 id=1002 data-nosnippet>1002</a> <span class="doccomment">/// Executes the default flow evaluation. Caller owns the execution guard.
|
|
<a href=#1003 id=1003 data-nosnippet>1003</a> </span><span class="kw">async fn </span>execute_once_default_unlocked(
|
|
<a href=#1004 id=1004 data-nosnippet>1004</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1005 id=1005 data-nosnippet>1005</a> engine: <span class="kw-2">&</span>QueryEngineRef,
|
|
<a href=#1006 id=1006 data-nosnippet>1006</a> frontend_client: <span class="kw-2">&</span>Arc<FrontendClient>,
|
|
<a href=#1007 id=1007 data-nosnippet>1007</a> max_window_cnt: <span class="prelude-ty">Option</span><usize>,
|
|
<a href=#1008 id=1008 data-nosnippet>1008</a> ) -> ExecuteOnceOutcome {
|
|
<a href=#1009 id=1009 data-nosnippet>1009</a> <span class="kw">let </span>new_query = <span class="kw">match </span><span class="self">self</span>.gen_insert_plan_unlocked(engine, max_window_cnt).<span class="kw">await </span>{
|
|
<a href=#1010 id=1010 data-nosnippet>1010</a> <span class="prelude-val">Ok</span>(new_query) => new_query,
|
|
<a href=#1011 id=1011 data-nosnippet>1011</a> <span class="prelude-val">Err</span>(err) => {
|
|
<a href=#1012 id=1012 data-nosnippet>1012</a> <span class="kw">return </span>ExecuteOnceOutcome {
|
|
<a href=#1013 id=1013 data-nosnippet>1013</a> new_query: <span class="prelude-val">None</span>,
|
|
<a href=#1014 id=1014 data-nosnippet>1014</a> result: <span class="prelude-val">Err</span>(err),
|
|
<a href=#1015 id=1015 data-nosnippet>1015</a> };
|
|
<a href=#1016 id=1016 data-nosnippet>1016</a> }
|
|
<a href=#1017 id=1017 data-nosnippet>1017</a> };
|
|
<a href=#1018 id=1018 data-nosnippet>1018</a>
|
|
<a href=#1019 id=1019 data-nosnippet>1019</a> <span class="kw">if let </span><span class="prelude-val">Some</span>(new_query) = new_query {
|
|
<a href=#1020 id=1020 data-nosnippet>1020</a> <span class="macro">debug!</span>(<span class="string">"Generate new query: {}"</span>, new_query.plan);
|
|
<a href=#1021 id=1021 data-nosnippet>1021</a> <span class="kw">let </span>res = <span class="self">self
|
|
<a href=#1022 id=1022 data-nosnippet>1022</a> </span>.execute_logical_plan_unlocked(
|
|
<a href=#1023 id=1023 data-nosnippet>1023</a> engine,
|
|
<a href=#1024 id=1024 data-nosnippet>1024</a> frontend_client,
|
|
<a href=#1025 id=1025 data-nosnippet>1025</a> <span class="kw-2">&</span>new_query.plan,
|
|
<a href=#1026 id=1026 data-nosnippet>1026</a> <span class="kw-2">&</span>new_query.dirty_restore,
|
|
<a href=#1027 id=1027 data-nosnippet>1027</a> <span class="kw-2">&</span>new_query.coverage,
|
|
<a href=#1028 id=1028 data-nosnippet>1028</a> )
|
|
<a href=#1029 id=1029 data-nosnippet>1029</a> .<span class="kw">await</span>;
|
|
<a href=#1030 id=1030 data-nosnippet>1030</a> <span class="kw">if </span>res.is_err() {
|
|
<a href=#1031 id=1031 data-nosnippet>1031</a> <span class="self">self</span>.handle_executed_query_failure(<span class="prelude-val">Some</span>(<span class="kw-2">&</span>new_query));
|
|
<a href=#1032 id=1032 data-nosnippet>1032</a> }
|
|
<a href=#1033 id=1033 data-nosnippet>1033</a> ExecuteOnceOutcome {
|
|
<a href=#1034 id=1034 data-nosnippet>1034</a> new_query: <span class="prelude-val">Some</span>(new_query),
|
|
<a href=#1035 id=1035 data-nosnippet>1035</a> result: res,
|
|
<a href=#1036 id=1036 data-nosnippet>1036</a> }
|
|
<a href=#1037 id=1037 data-nosnippet>1037</a> } <span class="kw">else </span>{
|
|
<a href=#1038 id=1038 data-nosnippet>1038</a> <span class="macro">debug!</span>(<span class="string">"Generate no query"</span>);
|
|
<a href=#1039 id=1039 data-nosnippet>1039</a> ExecuteOnceOutcome {
|
|
<a href=#1040 id=1040 data-nosnippet>1040</a> new_query: <span class="prelude-val">None</span>,
|
|
<a href=#1041 id=1041 data-nosnippet>1041</a> result: <span class="prelude-val">Ok</span>(<span class="prelude-val">None</span>),
|
|
<a href=#1042 id=1042 data-nosnippet>1042</a> }
|
|
<a href=#1043 id=1043 data-nosnippet>1043</a> }
|
|
<a href=#1044 id=1044 data-nosnippet>1044</a> }
|
|
<a href=#1045 id=1045 data-nosnippet>1045</a>
|
|
<a href=#1046 id=1046 data-nosnippet>1046</a> <span class="doccomment">/// Generates the ordinary insert plan. Caller must reach this through the serialized path.
|
|
<a href=#1047 id=1047 data-nosnippet>1047</a> </span><span class="kw">async fn </span>gen_insert_plan_unlocked(
|
|
<a href=#1048 id=1048 data-nosnippet>1048</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1049 id=1049 data-nosnippet>1049</a> engine: <span class="kw-2">&</span>QueryEngineRef,
|
|
<a href=#1050 id=1050 data-nosnippet>1050</a> max_window_cnt: <span class="prelude-ty">Option</span><usize>,
|
|
<a href=#1051 id=1051 data-nosnippet>1051</a> ) -> <span class="prelude-ty">Result</span><<span class="prelude-ty">Option</span><PlanInfo>, Error> {
|
|
<a href=#1052 id=1052 data-nosnippet>1052</a> <span class="self">self</span>.gen_insert_plan_with_values_unlocked(engine, max_window_cnt, <span class="kw-2">&</span>BTreeMap::new(), <span class="bool-val">false</span>)
|
|
<a href=#1053 id=1053 data-nosnippet>1053</a> .<span class="kw">await
|
|
<a href=#1054 id=1054 data-nosnippet>1054</a> </span>}
|
|
<a href=#1055 id=1055 data-nosnippet>1055</a>
|
|
<a href=#1056 id=1056 data-nosnippet>1056</a> <span class="doccomment">/// Generates the insert plan with extension-owned typed sink values. Caller must own the
|
|
<a href=#1057 id=1057 data-nosnippet>1057</a> /// serialized execution guard.
|
|
<a href=#1058 id=1058 data-nosnippet>1058</a> </span><span class="kw">pub async fn </span>gen_insert_plan_with_values_unlocked(
|
|
<a href=#1059 id=1059 data-nosnippet>1059</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1060 id=1060 data-nosnippet>1060</a> engine: <span class="kw-2">&</span>QueryEngineRef,
|
|
<a href=#1061 id=1061 data-nosnippet>1061</a> max_window_cnt: <span class="prelude-ty">Option</span><usize>,
|
|
<a href=#1062 id=1062 data-nosnippet>1062</a> values: <span class="kw-2">&</span>BTreeMap<String, ScalarValue>,
|
|
<a href=#1063 id=1063 data-nosnippet>1063</a> force_full_snapshot: bool,
|
|
<a href=#1064 id=1064 data-nosnippet>1064</a> ) -> <span class="prelude-ty">Result</span><<span class="prelude-ty">Option</span><PlanInfo>, Error> {
|
|
<a href=#1065 id=1065 data-nosnippet>1065</a> <span class="kw">let </span>(table, df_schema) = get_table_info_df_schema(
|
|
<a href=#1066 id=1066 data-nosnippet>1066</a> <span class="self">self</span>.config.catalog_manager.clone(),
|
|
<a href=#1067 id=1067 data-nosnippet>1067</a> <span class="self">self</span>.config.sink_table_name.clone(),
|
|
<a href=#1068 id=1068 data-nosnippet>1068</a> )
|
|
<a href=#1069 id=1069 data-nosnippet>1069</a> .<span class="kw">await</span><span class="question-mark">?</span>;
|
|
<a href=#1070 id=1070 data-nosnippet>1070</a>
|
|
<a href=#1071 id=1071 data-nosnippet>1071</a> <span class="kw">let </span>table_meta = <span class="kw-2">&</span>table.table_info().meta;
|
|
<a href=#1072 id=1072 data-nosnippet>1072</a> <span class="kw">let </span>merge_mode_last_non_null =
|
|
<a href=#1073 id=1073 data-nosnippet>1073</a> is_merge_mode_last_non_null(<span class="kw-2">&</span>table_meta.options.extra_options);
|
|
<a href=#1074 id=1074 data-nosnippet>1074</a> <span class="kw">let </span>primary_key_indices = table_meta.primary_key_indices.clone();
|
|
<a href=#1075 id=1075 data-nosnippet>1075</a>
|
|
<a href=#1076 id=1076 data-nosnippet>1076</a> <span class="kw">let </span>new_query = <span class="self">self
|
|
<a href=#1077 id=1077 data-nosnippet>1077</a> </span>.gen_query_with_time_window_with_values(
|
|
<a href=#1078 id=1078 data-nosnippet>1078</a> engine.clone(),
|
|
<a href=#1079 id=1079 data-nosnippet>1079</a> <span class="kw-2">&</span>table.table_info().meta.schema,
|
|
<a href=#1080 id=1080 data-nosnippet>1080</a> <span class="kw-2">&</span>primary_key_indices,
|
|
<a href=#1081 id=1081 data-nosnippet>1081</a> merge_mode_last_non_null,
|
|
<a href=#1082 id=1082 data-nosnippet>1082</a> max_window_cnt,
|
|
<a href=#1083 id=1083 data-nosnippet>1083</a> values,
|
|
<a href=#1084 id=1084 data-nosnippet>1084</a> force_full_snapshot,
|
|
<a href=#1085 id=1085 data-nosnippet>1085</a> )
|
|
<a href=#1086 id=1086 data-nosnippet>1086</a> .<span class="kw">await</span><span class="question-mark">?</span>;
|
|
<a href=#1087 id=1087 data-nosnippet>1087</a>
|
|
<a href=#1088 id=1088 data-nosnippet>1088</a> <span class="kw">let </span><span class="prelude-val">Some</span>(new_query) = new_query <span class="kw">else </span>{
|
|
<a href=#1089 id=1089 data-nosnippet>1089</a> <span class="kw">return </span><span class="prelude-val">Ok</span>(<span class="prelude-val">None</span>);
|
|
<a href=#1090 id=1090 data-nosnippet>1090</a> };
|
|
<a href=#1091 id=1091 data-nosnippet>1091</a>
|
|
<a href=#1092 id=1092 data-nosnippet>1092</a> <span class="comment">// first check if all columns in input query exists in sink table
|
|
<a href=#1093 id=1093 data-nosnippet>1093</a> // since insert into ref to names in record batch generate by given query
|
|
<a href=#1094 id=1094 data-nosnippet>1094</a> </span><span class="kw">let </span>table_columns = df_schema
|
|
<a href=#1095 id=1095 data-nosnippet>1095</a> .columns()
|
|
<a href=#1096 id=1096 data-nosnippet>1096</a> .into_iter()
|
|
<a href=#1097 id=1097 data-nosnippet>1097</a> .map(|c| c.name)
|
|
<a href=#1098 id=1098 data-nosnippet>1098</a> .collect::<BTreeSet<<span class="kw">_</span>>>();
|
|
<a href=#1099 id=1099 data-nosnippet>1099</a> <span class="kw">for </span>column <span class="kw">in </span>new_query.plan.schema().columns() {
|
|
<a href=#1100 id=1100 data-nosnippet>1100</a> <span class="kw">if </span>!table_columns.contains(column.name()) {
|
|
<a href=#1101 id=1101 data-nosnippet>1101</a> <span class="self">self</span>.restore_dirty_windows_after_failure(<span class="kw-2">&</span>new_query);
|
|
<a href=#1102 id=1102 data-nosnippet>1102</a> <span class="kw">return </span>InvalidQuerySnafu {
|
|
<a href=#1103 id=1103 data-nosnippet>1103</a> reason: <span class="macro">format!</span>(
|
|
<a href=#1104 id=1104 data-nosnippet>1104</a> <span class="string">"Column {} not found in sink table with columns {:?}"</span>,
|
|
<a href=#1105 id=1105 data-nosnippet>1105</a> column, table_columns
|
|
<a href=#1106 id=1106 data-nosnippet>1106</a> ),
|
|
<a href=#1107 id=1107 data-nosnippet>1107</a> }
|
|
<a href=#1108 id=1108 data-nosnippet>1108</a> .fail();
|
|
<a href=#1109 id=1109 data-nosnippet>1109</a> }
|
|
<a href=#1110 id=1110 data-nosnippet>1110</a> }
|
|
<a href=#1111 id=1111 data-nosnippet>1111</a>
|
|
<a href=#1112 id=1112 data-nosnippet>1112</a> <span class="kw">let </span>table_provider = Arc::new(DfTableProviderAdapter::new(table));
|
|
<a href=#1113 id=1113 data-nosnippet>1113</a> <span class="kw">let </span>table_source = Arc::new(DefaultTableSource::new(table_provider));
|
|
<a href=#1114 id=1114 data-nosnippet>1114</a>
|
|
<a href=#1115 id=1115 data-nosnippet>1115</a> <span class="comment">// update_at& time index placeholder (if exists) should have default value
|
|
<a href=#1116 id=1116 data-nosnippet>1116</a> </span><span class="kw">let </span>plan = LogicalPlan::Dml(DmlStatement::new(
|
|
<a href=#1117 id=1117 data-nosnippet>1117</a> datafusion_common::TableReference::Full {
|
|
<a href=#1118 id=1118 data-nosnippet>1118</a> catalog: <span class="self">self</span>.config.sink_table_name[<span class="number">0</span>].clone().into(),
|
|
<a href=#1119 id=1119 data-nosnippet>1119</a> schema: <span class="self">self</span>.config.sink_table_name[<span class="number">1</span>].clone().into(),
|
|
<a href=#1120 id=1120 data-nosnippet>1120</a> table: <span class="self">self</span>.config.sink_table_name[<span class="number">2</span>].clone().into(),
|
|
<a href=#1121 id=1121 data-nosnippet>1121</a> },
|
|
<a href=#1122 id=1122 data-nosnippet>1122</a> table_source,
|
|
<a href=#1123 id=1123 data-nosnippet>1123</a> WriteOp::Insert(datafusion_expr::dml::InsertOp::Append),
|
|
<a href=#1124 id=1124 data-nosnippet>1124</a> Arc::new(new_query.plan.clone()),
|
|
<a href=#1125 id=1125 data-nosnippet>1125</a> ));
|
|
<a href=#1126 id=1126 data-nosnippet>1126</a> <span class="kw">let </span>insert_into_info = PlanInfo {
|
|
<a href=#1127 id=1127 data-nosnippet>1127</a> plan,
|
|
<a href=#1128 id=1128 data-nosnippet>1128</a> dirty_restore: new_query.dirty_restore,
|
|
<a href=#1129 id=1129 data-nosnippet>1129</a> coverage: new_query.coverage,
|
|
<a href=#1130 id=1130 data-nosnippet>1130</a> };
|
|
<a href=#1131 id=1131 data-nosnippet>1131</a> <span class="kw">let </span>insert_into =
|
|
<a href=#1132 id=1132 data-nosnippet>1132</a> <span class="kw">match </span>insert_into_info
|
|
<a href=#1133 id=1133 data-nosnippet>1133</a> .plan
|
|
<a href=#1134 id=1134 data-nosnippet>1134</a> .clone()
|
|
<a href=#1135 id=1135 data-nosnippet>1135</a> .recompute_schema()
|
|
<a href=#1136 id=1136 data-nosnippet>1136</a> .context(DatafusionSnafu {
|
|
<a href=#1137 id=1137 data-nosnippet>1137</a> context: <span class="string">"Failed to recompute schema"</span>,
|
|
<a href=#1138 id=1138 data-nosnippet>1138</a> }) {
|
|
<a href=#1139 id=1139 data-nosnippet>1139</a> <span class="prelude-val">Ok</span>(insert_into) => insert_into,
|
|
<a href=#1140 id=1140 data-nosnippet>1140</a> <span class="prelude-val">Err</span>(err) => {
|
|
<a href=#1141 id=1141 data-nosnippet>1141</a> <span class="self">self</span>.restore_dirty_windows_after_failure(<span class="kw-2">&</span>insert_into_info);
|
|
<a href=#1142 id=1142 data-nosnippet>1142</a> <span class="kw">return </span><span class="prelude-val">Err</span>(err);
|
|
<a href=#1143 id=1143 data-nosnippet>1143</a> }
|
|
<a href=#1144 id=1144 data-nosnippet>1144</a> };
|
|
<a href=#1145 id=1145 data-nosnippet>1145</a>
|
|
<a href=#1146 id=1146 data-nosnippet>1146</a> <span class="prelude-val">Ok</span>(<span class="prelude-val">Some</span>(PlanInfo {
|
|
<a href=#1147 id=1147 data-nosnippet>1147</a> plan: insert_into,
|
|
<a href=#1148 id=1148 data-nosnippet>1148</a> dirty_restore: insert_into_info.dirty_restore,
|
|
<a href=#1149 id=1149 data-nosnippet>1149</a> coverage: insert_into_info.coverage,
|
|
<a href=#1150 id=1150 data-nosnippet>1150</a> }))
|
|
<a href=#1151 id=1151 data-nosnippet>1151</a> }
|
|
<a href=#1152 id=1152 data-nosnippet>1152</a>
|
|
<a href=#1153 id=1153 data-nosnippet>1153</a> <span class="kw">pub async fn </span>create_table(
|
|
<a href=#1154 id=1154 data-nosnippet>1154</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1155 id=1155 data-nosnippet>1155</a> frontend_client: <span class="kw-2">&</span>Arc<FrontendClient>,
|
|
<a href=#1156 id=1156 data-nosnippet>1156</a> expr: CreateTableExpr,
|
|
<a href=#1157 id=1157 data-nosnippet>1157</a> ) -> <span class="prelude-ty">Result</span><(), Error> {
|
|
<a href=#1158 id=1158 data-nosnippet>1158</a> <span class="kw">let </span>catalog = <span class="kw-2">&</span><span class="self">self</span>.config.sink_table_name[<span class="number">0</span>];
|
|
<a href=#1159 id=1159 data-nosnippet>1159</a> <span class="kw">let </span>schema = <span class="kw-2">&</span><span class="self">self</span>.config.sink_table_name[<span class="number">1</span>];
|
|
<a href=#1160 id=1160 data-nosnippet>1160</a> frontend_client
|
|
<a href=#1161 id=1161 data-nosnippet>1161</a> .create(expr.clone(), catalog, schema)
|
|
<a href=#1162 id=1162 data-nosnippet>1162</a> .<span class="kw">await</span><span class="question-mark">?</span>;
|
|
<a href=#1163 id=1163 data-nosnippet>1163</a> <span class="prelude-val">Ok</span>(())
|
|
<a href=#1164 id=1164 data-nosnippet>1164</a> }
|
|
<a href=#1165 id=1165 data-nosnippet>1165</a>
|
|
<a href=#1166 id=1166 data-nosnippet>1166</a> <span class="doccomment">/// Applies checkpoint state updates around the raw insert-plan execution.
|
|
<a href=#1167 id=1167 data-nosnippet>1167</a> </span><span class="kw">async fn </span>execute_logical_plan_unlocked(
|
|
<a href=#1168 id=1168 data-nosnippet>1168</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1169 id=1169 data-nosnippet>1169</a> engine: <span class="kw-2">&</span>QueryEngineRef,
|
|
<a href=#1170 id=1170 data-nosnippet>1170</a> frontend_client: <span class="kw-2">&</span>Arc<FrontendClient>,
|
|
<a href=#1171 id=1171 data-nosnippet>1171</a> plan: <span class="kw-2">&</span>LogicalPlan,
|
|
<a href=#1172 id=1172 data-nosnippet>1172</a> dirty_restore: <span class="kw-2">&</span>DirtyRestore,
|
|
<a href=#1173 id=1173 data-nosnippet>1173</a> coverage: <span class="kw-2">&</span>QueryCoverage,
|
|
<a href=#1174 id=1174 data-nosnippet>1174</a> ) -> <span class="prelude-ty">Result</span><<span class="prelude-ty">Option</span><(usize, Duration)>, Error> {
|
|
<a href=#1175 id=1175 data-nosnippet>1175</a> <span class="kw">let </span>flow_id = <span class="self">self</span>.config.flow_id;
|
|
<a href=#1176 id=1176 data-nosnippet>1176</a> <span class="kw">let </span><span class="prelude-val">Some</span>((res, elapsed)) = <span class="self">self
|
|
<a href=#1177 id=1177 data-nosnippet>1177</a> </span>.execute_plan_unlocked(engine, frontend_client, plan, dirty_restore, coverage)
|
|
<a href=#1178 id=1178 data-nosnippet>1178</a> .<span class="kw">await</span><span class="question-mark">?
|
|
<a href=#1179 id=1179 data-nosnippet>1179</a> </span><span class="kw">else </span>{
|
|
<a href=#1180 id=1180 data-nosnippet>1180</a> <span class="kw">return </span><span class="prelude-val">Ok</span>(<span class="prelude-val">None</span>);
|
|
<a href=#1181 id=1181 data-nosnippet>1181</a> };
|
|
<a href=#1182 id=1182 data-nosnippet>1182</a>
|
|
<a href=#1183 id=1183 data-nosnippet>1183</a> <span class="kw">if let </span><span class="prelude-val">Err</span>(err) = <span class="kw-2">&</span>res {
|
|
<a href=#1184 id=1184 data-nosnippet>1184</a> <span class="kw">let </span>decision = {
|
|
<a href=#1185 id=1185 data-nosnippet>1185</a> <span class="kw">let </span><span class="kw-2">mut </span>state = <span class="self">self</span>.state.write().unwrap();
|
|
<a href=#1186 id=1186 data-nosnippet>1186</a> <span class="kw">let </span>reason = <span class="self">Self</span>::query_failure_reason(err, coverage);
|
|
<a href=#1187 id=1187 data-nosnippet>1187</a> <span class="self">Self</span>::apply_query_failure_to_state(<span class="kw-2">&mut </span>state, elapsed, coverage, reason)
|
|
<a href=#1188 id=1188 data-nosnippet>1188</a> };
|
|
<a href=#1189 id=1189 data-nosnippet>1189</a> <span class="kw">if let </span><span class="prelude-val">Some</span>(decision) = decision {
|
|
<a href=#1190 id=1190 data-nosnippet>1190</a> <span class="self">Self</span>::record_checkpoint_decision(flow_id, decision);
|
|
<a href=#1191 id=1191 data-nosnippet>1191</a> }
|
|
<a href=#1192 id=1192 data-nosnippet>1192</a> }
|
|
<a href=#1193 id=1193 data-nosnippet>1193</a>
|
|
<a href=#1194 id=1194 data-nosnippet>1194</a> <span class="kw">let </span>res = res<span class="question-mark">?</span>;
|
|
<a href=#1195 id=1195 data-nosnippet>1195</a> <span class="kw">let </span>(affected_rows, <span class="kw">_</span>) = res.output.extract_rows_and_cost();
|
|
<a href=#1196 id=1196 data-nosnippet>1196</a> <span class="kw">let </span>decision = {
|
|
<a href=#1197 id=1197 data-nosnippet>1197</a> <span class="kw">let </span><span class="kw-2">mut </span>state = <span class="self">self</span>.state.write().unwrap();
|
|
<a href=#1198 id=1198 data-nosnippet>1198</a> <span class="self">Self</span>::apply_query_result_to_state(<span class="kw-2">&mut </span>state, <span class="kw-2">&</span>res, elapsed, coverage)
|
|
<a href=#1199 id=1199 data-nosnippet>1199</a> };
|
|
<a href=#1200 id=1200 data-nosnippet>1200</a> <span class="self">Self</span>::record_checkpoint_decision(flow_id, decision);
|
|
<a href=#1201 id=1201 data-nosnippet>1201</a> <span class="prelude-val">Ok</span>(<span class="prelude-val">Some</span>((affected_rows, elapsed)))
|
|
<a href=#1202 id=1202 data-nosnippet>1202</a> }
|
|
<a href=#1203 id=1203 data-nosnippet>1203</a>
|
|
<a href=#1204 id=1204 data-nosnippet>1204</a> <span class="doccomment">/// Executes a fully constructed insert plan without checkpoint-state mutation.
|
|
<a href=#1205 id=1205 data-nosnippet>1205</a> /// Callers must own the serialized execution guard. Its unsafe-incremental safety fallback
|
|
<a href=#1206 id=1206 data-nosnippet>1206</a> /// still restores the supplied dirty work before returning `Ok(None)`.
|
|
<a href=#1207 id=1207 data-nosnippet>1207</a> </span><span class="kw">pub async fn </span>execute_plan_unlocked(
|
|
<a href=#1208 id=1208 data-nosnippet>1208</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1209 id=1209 data-nosnippet>1209</a> engine: <span class="kw-2">&</span>QueryEngineRef,
|
|
<a href=#1210 id=1210 data-nosnippet>1210</a> frontend_client: <span class="kw-2">&</span>Arc<FrontendClient>,
|
|
<a href=#1211 id=1211 data-nosnippet>1211</a> plan: <span class="kw-2">&</span>LogicalPlan,
|
|
<a href=#1212 id=1212 data-nosnippet>1212</a> dirty_restore: <span class="kw-2">&</span>DirtyRestore,
|
|
<a href=#1213 id=1213 data-nosnippet>1213</a> coverage: <span class="kw-2">&</span>QueryCoverage,
|
|
<a href=#1214 id=1214 data-nosnippet>1214</a> ) -> <span class="prelude-ty">Result</span><<span class="prelude-ty">Option</span><(<span class="prelude-ty">Result</span><OutputWithMetrics, Error>, Duration)>, Error> {
|
|
<a href=#1215 id=1215 data-nosnippet>1215</a> <span class="kw">let </span>instant = Instant::now();
|
|
<a href=#1216 id=1216 data-nosnippet>1216</a> <span class="kw">let </span>flow_id = <span class="self">self</span>.config.flow_id;
|
|
<a href=#1217 id=1217 data-nosnippet>1217</a>
|
|
<a href=#1218 id=1218 data-nosnippet>1218</a> <span class="macro">debug!</span>(
|
|
<a href=#1219 id=1219 data-nosnippet>1219</a> <span class="string">"Executing flow {flow_id}(expire_after={:?} secs) with query {}"</span>,
|
|
<a href=#1220 id=1220 data-nosnippet>1220</a> <span class="self">self</span>.config.expire_after, <span class="kw-2">&</span>plan
|
|
<a href=#1221 id=1221 data-nosnippet>1221</a> );
|
|
<a href=#1222 id=1222 data-nosnippet>1222</a>
|
|
<a href=#1223 id=1223 data-nosnippet>1223</a> <span class="kw">let </span>catalog = <span class="kw-2">&</span><span class="self">self</span>.config.sink_table_name[<span class="number">0</span>];
|
|
<a href=#1224 id=1224 data-nosnippet>1224</a> <span class="kw">let </span>schema = <span class="kw-2">&</span><span class="self">self</span>.config.sink_table_name[<span class="number">1</span>];
|
|
<a href=#1225 id=1225 data-nosnippet>1225</a>
|
|
<a href=#1226 id=1226 data-nosnippet>1226</a> <span class="comment">// fix all table ref by make it fully qualified, i.e. "table_name" => "catalog_name.schema_name.table_name"
|
|
<a href=#1227 id=1227 data-nosnippet>1227</a> </span><span class="kw">let </span>plan = plan
|
|
<a href=#1228 id=1228 data-nosnippet>1228</a> .clone()
|
|
<a href=#1229 id=1229 data-nosnippet>1229</a> .transform_down_with_subqueries(|p| {
|
|
<a href=#1230 id=1230 data-nosnippet>1230</a> <span class="kw">if let </span>LogicalPlan::TableScan(<span class="kw-2">mut </span>table_scan) = p {
|
|
<a href=#1231 id=1231 data-nosnippet>1231</a> <span class="kw">let </span>resolved = table_scan.table_name.resolve(catalog, schema);
|
|
<a href=#1232 id=1232 data-nosnippet>1232</a> table_scan.table_name = resolved.into();
|
|
<a href=#1233 id=1233 data-nosnippet>1233</a> <span class="prelude-val">Ok</span>(Transformed::yes(LogicalPlan::TableScan(table_scan)))
|
|
<a href=#1234 id=1234 data-nosnippet>1234</a> } <span class="kw">else </span>{
|
|
<a href=#1235 id=1235 data-nosnippet>1235</a> <span class="prelude-val">Ok</span>(Transformed::no(p))
|
|
<a href=#1236 id=1236 data-nosnippet>1236</a> }
|
|
<a href=#1237 id=1237 data-nosnippet>1237</a> })
|
|
<a href=#1238 id=1238 data-nosnippet>1238</a> .with_context(|<span class="kw">_</span>| DatafusionSnafu {
|
|
<a href=#1239 id=1239 data-nosnippet>1239</a> context: <span class="macro">format!</span>(<span class="string">"Failed to fix table ref in logical plan, plan={:?}"</span>, plan),
|
|
<a href=#1240 id=1240 data-nosnippet>1240</a> })<span class="question-mark">?
|
|
<a href=#1241 id=1241 data-nosnippet>1241</a> </span>.data;
|
|
<a href=#1242 id=1242 data-nosnippet>1242</a>
|
|
<a href=#1243 id=1243 data-nosnippet>1243</a> <span class="comment">// For incremental-mode SQL queries, attempt to rewrite the delta aggregate
|
|
<a href=#1244 id=1244 data-nosnippet>1244</a> // plan into a safe delta-LEFT-JOIN-sink form before deciding on extensions.
|
|
<a href=#1245 id=1245 data-nosnippet>1245</a> </span><span class="kw">let </span>incremental_plan = <span class="kw">if </span>coverage.is_incremental_delta() {
|
|
<a href=#1246 id=1246 data-nosnippet>1246</a> <span class="self">self</span>.prepare_plan_for_incremental(engine, <span class="kw-2">&</span>plan).<span class="kw">await</span><span class="question-mark">?
|
|
<a href=#1247 id=1247 data-nosnippet>1247</a> </span>} <span class="kw">else </span>{
|
|
<a href=#1248 id=1248 data-nosnippet>1248</a> <span class="prelude-val">None
|
|
<a href=#1249 id=1249 data-nosnippet>1249</a> </span>};
|
|
<a href=#1250 id=1250 data-nosnippet>1250</a> <span class="kw">let </span>incremental_safe = incremental_plan.is_some();
|
|
<a href=#1251 id=1251 data-nosnippet>1251</a> <span class="kw">if </span>coverage.is_incremental_delta() && !incremental_safe {
|
|
<a href=#1252 id=1252 data-nosnippet>1252</a> <span class="macro">debug!</span>(
|
|
<a href=#1253 id=1253 data-nosnippet>1253</a> <span class="string">"Flow {flow_id} skipped unsafe incremental delta fallback; \
|
|
<a href=#1254 id=1254 data-nosnippet>1254</a> restored dirty signal instead of executing an unfiltered full snapshot"
|
|
<a href=#1255 id=1255 data-nosnippet>1255</a> </span>);
|
|
<a href=#1256 id=1256 data-nosnippet>1256</a> <span class="self">self</span>.restore_dirty_windows(dirty_restore);
|
|
<a href=#1257 id=1257 data-nosnippet>1257</a> <span class="kw">return </span><span class="prelude-val">Ok</span>(<span class="prelude-val">None</span>);
|
|
<a href=#1258 id=1258 data-nosnippet>1258</a> }
|
|
<a href=#1259 id=1259 data-nosnippet>1259</a> <span class="kw">let </span>plan = incremental_plan.unwrap_or_else(|| plan.clone());
|
|
<a href=#1260 id=1260 data-nosnippet>1260</a> <span class="kw">let </span>plan = <span class="kw">match </span><span class="kw-2">&</span><span class="self">self</span>.execution {
|
|
<a href=#1261 id=1261 data-nosnippet>1261</a> <span class="prelude-val">Some</span>(execution) => execution.rewrite_plan(<span class="self">self</span>, plan)<span class="question-mark">?</span>,
|
|
<a href=#1262 id=1262 data-nosnippet>1262</a> <span class="prelude-val">None </span>=> plan,
|
|
<a href=#1263 id=1263 data-nosnippet>1263</a> };
|
|
<a href=#1264 id=1264 data-nosnippet>1264</a>
|
|
<a href=#1265 id=1265 data-nosnippet>1265</a> <span class="kw">let </span>extensions = <span class="self">self
|
|
<a href=#1266 id=1266 data-nosnippet>1266</a> </span>.build_flow_query_extensions(incremental_safe, coverage.is_incremental_delta())
|
|
<a href=#1267 id=1267 data-nosnippet>1267</a> .<span class="kw">await</span><span class="question-mark">?</span>;
|
|
<a href=#1268 id=1268 data-nosnippet>1268</a> <span class="kw">let </span>frontend_extensions = <span class="self">self</span>.frontend_extensions();
|
|
<a href=#1269 id=1269 data-nosnippet>1269</a> <span class="kw">let </span>extension_refs = extensions
|
|
<a href=#1270 id=1270 data-nosnippet>1270</a> .iter()
|
|
<a href=#1271 id=1271 data-nosnippet>1271</a> .map(|(key, value)| (<span class="kw-2">*</span>key, value.as_str()))
|
|
<a href=#1272 id=1272 data-nosnippet>1272</a> .chain(
|
|
<a href=#1273 id=1273 data-nosnippet>1273</a> frontend_extensions
|
|
<a href=#1274 id=1274 data-nosnippet>1274</a> .iter()
|
|
<a href=#1275 id=1275 data-nosnippet>1275</a> .map(|(key, value)| (key.as_str(), value.as_str())),
|
|
<a href=#1276 id=1276 data-nosnippet>1276</a> )
|
|
<a href=#1277 id=1277 data-nosnippet>1277</a> .collect::<Vec<<span class="kw">_</span>>>();
|
|
<a href=#1278 id=1278 data-nosnippet>1278</a> <span class="kw">let </span>query_mode = <span class="kw">if </span>extensions
|
|
<a href=#1279 id=1279 data-nosnippet>1279</a> .iter()
|
|
<a href=#1280 id=1280 data-nosnippet>1280</a> .any(|(key, <span class="kw">_</span>)| <span class="kw-2">*</span>key == FLOW_INCREMENTAL_MODE)
|
|
<a href=#1281 id=1281 data-nosnippet>1281</a> {
|
|
<a href=#1282 id=1282 data-nosnippet>1282</a> CheckpointMode::Incremental
|
|
<a href=#1283 id=1283 data-nosnippet>1283</a> } <span class="kw">else </span>{
|
|
<a href=#1284 id=1284 data-nosnippet>1284</a> CheckpointMode::FullSnapshot
|
|
<a href=#1285 id=1285 data-nosnippet>1285</a> };
|
|
<a href=#1286 id=1286 data-nosnippet>1286</a> <span class="self">Self</span>::record_query_mode(flow_id, query_mode);
|
|
<a href=#1287 id=1287 data-nosnippet>1287</a> <span class="macro">debug!</span>(
|
|
<a href=#1288 id=1288 data-nosnippet>1288</a> <span class="string">"Flow {flow_id} executing batching query with checkpoint_mode={}, extension_count={}"</span>,
|
|
<a href=#1289 id=1289 data-nosnippet>1289</a> checkpoint_mode_label(query_mode),
|
|
<a href=#1290 id=1290 data-nosnippet>1290</a> extensions.len()
|
|
<a href=#1291 id=1291 data-nosnippet>1291</a> );
|
|
<a href=#1292 id=1292 data-nosnippet>1292</a>
|
|
<a href=#1293 id=1293 data-nosnippet>1293</a> <span class="kw">let </span><span class="kw-2">mut </span>peer_desc = <span class="prelude-val">None</span>;
|
|
<a href=#1294 id=1294 data-nosnippet>1294</a> <span class="kw">let </span>res = {
|
|
<a href=#1295 id=1295 data-nosnippet>1295</a> <span class="kw">let </span>_timer = METRIC_FLOW_BATCHING_ENGINE_QUERY_TIME
|
|
<a href=#1296 id=1296 data-nosnippet>1296</a> .with_label_values(<span class="kw-2">&</span>[flow_id.to_string().as_str()])
|
|
<a href=#1297 id=1297 data-nosnippet>1297</a> .start_timer();
|
|
<a href=#1298 id=1298 data-nosnippet>1298</a>
|
|
<a href=#1299 id=1299 data-nosnippet>1299</a> <span class="kw">let </span>req = <span class="kw">if let </span><span class="prelude-val">Some</span>((insert_to, insert_input_plan)) =
|
|
<a href=#1300 id=1300 data-nosnippet>1300</a> breakup_insert_plan(<span class="kw-2">&</span>plan, catalog, schema)
|
|
<a href=#1301 id=1301 data-nosnippet>1301</a> {
|
|
<a href=#1302 id=1302 data-nosnippet>1302</a> <span class="kw">if </span>query_mode == CheckpointMode::FullSnapshot
|
|
<a href=#1303 id=1303 data-nosnippet>1303</a> && <span class="macro">matches!</span>(<span class="self">self</span>.config.query_type, QueryType::Sql)
|
|
<a href=#1304 id=1304 data-nosnippet>1304</a> && <span class="self">self</span>.config.flow_eval_interval.is_some()
|
|
<a href=#1305 id=1305 data-nosnippet>1305</a> && <span class="self">self</span>.config.time_window_expr.is_none()
|
|
<a href=#1306 id=1306 data-nosnippet>1306</a> {
|
|
<a href=#1307 id=1307 data-nosnippet>1307</a> <span class="comment">// Evaluation-interval SQL flows without a time-window
|
|
<a href=#1308 id=1308 data-nosnippet>1308</a> // expression execute as full-query snapshots. Send these
|
|
<a href=#1309 id=1309 data-nosnippet>1309</a> // as SQL text instead of Substrait to avoid logical-plan
|
|
<a href=#1310 id=1310 data-nosnippet>1310</a> // round-trip issues around complex joins/unions/CTEs and
|
|
<a href=#1311 id=1311 data-nosnippet>1311</a> // duplicate field aliases. Keep ordinary SQL full snapshots
|
|
<a href=#1312 id=1312 data-nosnippet>1312</a> // on the existing InsertIntoPlan path because SQL unparsing
|
|
<a href=#1313 id=1313 data-nosnippet>1313</a> // is not valid for every planned aggregate shape yet.
|
|
<a href=#1314 id=1314 data-nosnippet>1314</a> // If the local SQL unparser does not support this plan,
|
|
<a href=#1315 id=1315 data-nosnippet>1315</a> // keep the previous InsertIntoPlan transport as a fallback.
|
|
<a href=#1316 id=1316 data-nosnippet>1316</a> </span><span class="kw">match </span>df_plan_to_sql(<span class="kw-2">&</span>insert_input_plan) {
|
|
<a href=#1317 id=1317 data-nosnippet>1317</a> <span class="prelude-val">Ok</span>(select_sql) => {
|
|
<a href=#1318 id=1318 data-nosnippet>1318</a> <span class="kw">let </span>target_columns = format_insert_target_columns(<span class="kw-2">&</span>insert_input_plan);
|
|
<a href=#1319 id=1319 data-nosnippet>1319</a> <span class="kw">let </span>sql = <span class="macro">format!</span>(
|
|
<a href=#1320 id=1320 data-nosnippet>1320</a> <span class="string">"INSERT INTO {} ({}) {}"</span>,
|
|
<a href=#1321 id=1321 data-nosnippet>1321</a> TableReference::full(
|
|
<a href=#1322 id=1322 data-nosnippet>1322</a> insert_to.catalog_name.as_str(),
|
|
<a href=#1323 id=1323 data-nosnippet>1323</a> insert_to.schema_name.as_str(),
|
|
<a href=#1324 id=1324 data-nosnippet>1324</a> insert_to.table_name.as_str(),
|
|
<a href=#1325 id=1325 data-nosnippet>1325</a> )
|
|
<a href=#1326 id=1326 data-nosnippet>1326</a> .to_quoted_string(),
|
|
<a href=#1327 id=1327 data-nosnippet>1327</a> target_columns,
|
|
<a href=#1328 id=1328 data-nosnippet>1328</a> select_sql
|
|
<a href=#1329 id=1329 data-nosnippet>1329</a> );
|
|
<a href=#1330 id=1330 data-nosnippet>1330</a> api::v1::QueryRequest {
|
|
<a href=#1331 id=1331 data-nosnippet>1331</a> query: <span class="prelude-val">Some</span>(api::v1::query_request::Query::Sql(sql)),
|
|
<a href=#1332 id=1332 data-nosnippet>1332</a> }
|
|
<a href=#1333 id=1333 data-nosnippet>1333</a> }
|
|
<a href=#1334 id=1334 data-nosnippet>1334</a> <span class="prelude-val">Err</span>(err) => {
|
|
<a href=#1335 id=1335 data-nosnippet>1335</a> <span class="macro">debug!</span>(
|
|
<a href=#1336 id=1336 data-nosnippet>1336</a> <span class="string">"Failed to unparse full-snapshot SQL flow {} plan; \
|
|
<a href=#1337 id=1337 data-nosnippet>1337</a> falling back to InsertIntoPlan: {:?}"</span>,
|
|
<a href=#1338 id=1338 data-nosnippet>1338</a> flow_id, err
|
|
<a href=#1339 id=1339 data-nosnippet>1339</a> );
|
|
<a href=#1340 id=1340 data-nosnippet>1340</a> encode_insert_plan_request(insert_to, <span class="kw-2">&</span>insert_input_plan)<span class="question-mark">?
|
|
<a href=#1341 id=1341 data-nosnippet>1341</a> </span>}
|
|
<a href=#1342 id=1342 data-nosnippet>1342</a> }
|
|
<a href=#1343 id=1343 data-nosnippet>1343</a> } <span class="kw">else </span>{
|
|
<a href=#1344 id=1344 data-nosnippet>1344</a> encode_insert_plan_request(insert_to, <span class="kw-2">&</span>insert_input_plan)<span class="question-mark">?
|
|
<a href=#1345 id=1345 data-nosnippet>1345</a> </span>}
|
|
<a href=#1346 id=1346 data-nosnippet>1346</a> } <span class="kw">else </span>{
|
|
<a href=#1347 id=1347 data-nosnippet>1347</a> <span class="kw">let </span>message = DFLogicalSubstraitConvertor {}
|
|
<a href=#1348 id=1348 data-nosnippet>1348</a> .encode(<span class="kw-2">&</span>plan, DefaultSerializer)
|
|
<a href=#1349 id=1349 data-nosnippet>1349</a> .context(SubstraitEncodeLogicalPlanSnafu)<span class="question-mark">?</span>;
|
|
<a href=#1350 id=1350 data-nosnippet>1350</a>
|
|
<a href=#1351 id=1351 data-nosnippet>1351</a> api::v1::QueryRequest {
|
|
<a href=#1352 id=1352 data-nosnippet>1352</a> query: <span class="prelude-val">Some</span>(api::v1::query_request::Query::LogicalPlan(message.to_vec())),
|
|
<a href=#1353 id=1353 data-nosnippet>1353</a> }
|
|
<a href=#1354 id=1354 data-nosnippet>1354</a> };
|
|
<a href=#1355 id=1355 data-nosnippet>1355</a>
|
|
<a href=#1356 id=1356 data-nosnippet>1356</a> <span class="kw">let </span>snapshot_seqs = coverage.snapshot_seqs();
|
|
<a href=#1357 id=1357 data-nosnippet>1357</a> {
|
|
<a href=#1358 id=1358 data-nosnippet>1358</a> <span class="kw">let </span><span class="kw-2">mut </span>state = <span class="self">self</span>.state.write().unwrap();
|
|
<a href=#1359 id=1359 data-nosnippet>1359</a> state.record_start_time_if_first();
|
|
<a href=#1360 id=1360 data-nosnippet>1360</a> }
|
|
<a href=#1361 id=1361 data-nosnippet>1361</a> frontend_client
|
|
<a href=#1362 id=1362 data-nosnippet>1362</a> .query_with_terminal_metrics(
|
|
<a href=#1363 id=1363 data-nosnippet>1363</a> catalog,
|
|
<a href=#1364 id=1364 data-nosnippet>1364</a> schema,
|
|
<a href=#1365 id=1365 data-nosnippet>1365</a> req,
|
|
<a href=#1366 id=1366 data-nosnippet>1366</a> <span class="kw-2">&</span>extension_refs,
|
|
<a href=#1367 id=1367 data-nosnippet>1367</a> <span class="kw-2">&</span>snapshot_seqs,
|
|
<a href=#1368 id=1368 data-nosnippet>1368</a> <span class="kw-2">&mut </span>peer_desc,
|
|
<a href=#1369 id=1369 data-nosnippet>1369</a> )
|
|
<a href=#1370 id=1370 data-nosnippet>1370</a> .<span class="kw">await
|
|
<a href=#1371 id=1371 data-nosnippet>1371</a> </span>};
|
|
<a href=#1372 id=1372 data-nosnippet>1372</a>
|
|
<a href=#1373 id=1373 data-nosnippet>1373</a> <span class="kw">let </span>elapsed = instant.elapsed();
|
|
<a href=#1374 id=1374 data-nosnippet>1374</a> <span class="kw">let </span>peer_label = peer_desc
|
|
<a href=#1375 id=1375 data-nosnippet>1375</a> .as_ref()
|
|
<a href=#1376 id=1376 data-nosnippet>1376</a> .map(ToString::to_string)
|
|
<a href=#1377 id=1377 data-nosnippet>1377</a> .unwrap_or_else(|| PeerDesc::default().to_string());
|
|
<a href=#1378 id=1378 data-nosnippet>1378</a> <span class="kw">if let </span><span class="prelude-val">Err</span>(err) = <span class="kw-2">&</span>res {
|
|
<a href=#1379 id=1379 data-nosnippet>1379</a> <span class="comment">// This raw helper deliberately leaves checkpoint transitions to its caller. The
|
|
<a href=#1380 id=1380 data-nosnippet>1380</a> // default wrapper performs the single OSS transition; extension owners can persist
|
|
<a href=#1381 id=1381 data-nosnippet>1381</a> // their durable marker before updating in-memory state.
|
|
<a href=#1382 id=1382 data-nosnippet>1382</a> </span><span class="macro">warn!</span>(
|
|
<a href=#1383 id=1383 data-nosnippet>1383</a> <span class="string">"Failed to execute Flow {flow_id} on frontend {peer_label}, result: {err:?}, elapsed: {:?} with query: {}"</span>,
|
|
<a href=#1384 id=1384 data-nosnippet>1384</a> elapsed, <span class="kw-2">&</span>plan
|
|
<a href=#1385 id=1385 data-nosnippet>1385</a> );
|
|
<a href=#1386 id=1386 data-nosnippet>1386</a> }
|
|
<a href=#1387 id=1387 data-nosnippet>1387</a>
|
|
<a href=#1388 id=1388 data-nosnippet>1388</a> <span class="comment">// record slow query
|
|
<a href=#1389 id=1389 data-nosnippet>1389</a> </span><span class="kw">if </span>elapsed >= <span class="self">self</span>.config.batch_opts.slow_query_threshold {
|
|
<a href=#1390 id=1390 data-nosnippet>1390</a> <span class="macro">warn!</span>(
|
|
<a href=#1391 id=1391 data-nosnippet>1391</a> <span class="string">"Flow {flow_id} on frontend {peer_label} executed for {:?} before complete, query: {}"</span>,
|
|
<a href=#1392 id=1392 data-nosnippet>1392</a> elapsed, <span class="kw-2">&</span>plan
|
|
<a href=#1393 id=1393 data-nosnippet>1393</a> );
|
|
<a href=#1394 id=1394 data-nosnippet>1394</a> <span class="kw">let </span>flow_id = flow_id.to_string();
|
|
<a href=#1395 id=1395 data-nosnippet>1395</a> METRIC_FLOW_BATCHING_ENGINE_SLOW_QUERY
|
|
<a href=#1396 id=1396 data-nosnippet>1396</a> .with_label_values(<span class="kw-2">&</span>[flow_id.as_str(), peer_label.as_str()])
|
|
<a href=#1397 id=1397 data-nosnippet>1397</a> .observe(elapsed.as_secs_f64());
|
|
<a href=#1398 id=1398 data-nosnippet>1398</a> }
|
|
<a href=#1399 id=1399 data-nosnippet>1399</a>
|
|
<a href=#1400 id=1400 data-nosnippet>1400</a> <span class="kw">match </span>res {
|
|
<a href=#1401 id=1401 data-nosnippet>1401</a> <span class="prelude-val">Ok</span>(res) => {
|
|
<a href=#1402 id=1402 data-nosnippet>1402</a> <span class="kw">let </span>(affected_rows, <span class="kw">_</span>) = res.output.extract_rows_and_cost();
|
|
<a href=#1403 id=1403 data-nosnippet>1403</a> <span class="kw">if </span><span class="macro">matches!</span>(<span class="kw-2">&</span>res.output.data, common_query::OutputData::AffectedRows(<span class="kw">_</span>)) {
|
|
<a href=#1404 id=1404 data-nosnippet>1404</a> res.metrics.wait_ready().<span class="kw">await</span>;
|
|
<a href=#1405 id=1405 data-nosnippet>1405</a> }
|
|
<a href=#1406 id=1406 data-nosnippet>1406</a> <span class="kw">if let </span><span class="prelude-val">Some</span>(error) = res.metrics.completion_error() {
|
|
<a href=#1407 id=1407 data-nosnippet>1407</a> <span class="macro">warn!</span>(<span class="string">"Flow {flow_id} completed with terminal metrics error: {error}"</span>);
|
|
<a href=#1408 id=1408 data-nosnippet>1408</a> }
|
|
<a href=#1409 id=1409 data-nosnippet>1409</a> <span class="macro">debug!</span>(
|
|
<a href=#1410 id=1410 data-nosnippet>1410</a> <span class="string">"Flow {flow_id} executed, affected_rows: {affected_rows:?}, elapsed: {:?}, watermark: {:?}"</span>,
|
|
<a href=#1411 id=1411 data-nosnippet>1411</a> elapsed,
|
|
<a href=#1412 id=1412 data-nosnippet>1412</a> res.region_watermark_map()
|
|
<a href=#1413 id=1413 data-nosnippet>1413</a> );
|
|
<a href=#1414 id=1414 data-nosnippet>1414</a> METRIC_FLOW_ROWS
|
|
<a href=#1415 id=1415 data-nosnippet>1415</a> .with_label_values(<span class="kw-2">&</span>[<span class="macro">format!</span>(<span class="string">"{}-out-batching"</span>, flow_id).as_str()])
|
|
<a href=#1416 id=1416 data-nosnippet>1416</a> .inc_by(affected_rows <span class="kw">as _</span>);
|
|
<a href=#1417 id=1417 data-nosnippet>1417</a> <span class="prelude-val">Ok</span>(<span class="prelude-val">Some</span>((<span class="prelude-val">Ok</span>(res), elapsed)))
|
|
<a href=#1418 id=1418 data-nosnippet>1418</a> }
|
|
<a href=#1419 id=1419 data-nosnippet>1419</a> <span class="prelude-val">Err</span>(err) => <span class="prelude-val">Ok</span>(<span class="prelude-val">Some</span>((<span class="prelude-val">Err</span>(err), elapsed))),
|
|
<a href=#1420 id=1420 data-nosnippet>1420</a> }
|
|
<a href=#1421 id=1421 data-nosnippet>1421</a> }
|
|
<a href=#1422 id=1422 data-nosnippet>1422</a>
|
|
<a href=#1423 id=1423 data-nosnippet>1423</a> <span class="doccomment">/// Restore dirty windows consumed by a failed query so they are retried on
|
|
<a href=#1424 id=1424 data-nosnippet>1424</a> /// the next execution.
|
|
<a href=#1425 id=1425 data-nosnippet>1425</a> ///
|
|
<a href=#1426 id=1426 data-nosnippet>1426</a> </span><span class="kw">pub fn </span>restore_dirty_windows(<span class="kw-2">&</span><span class="self">self</span>, dirty_restore: <span class="kw-2">&</span>DirtyRestore) {
|
|
<a href=#1427 id=1427 data-nosnippet>1427</a> <span class="kw">match </span>dirty_restore {
|
|
<a href=#1428 id=1428 data-nosnippet>1428</a> DirtyRestore::Scoped(filter) => <span class="self">self</span>.restore_scoped_dirty_windows(filter),
|
|
<a href=#1429 id=1429 data-nosnippet>1429</a> DirtyRestore::Unscoped(dirty_windows) => <span class="self">self
|
|
<a href=#1430 id=1430 data-nosnippet>1430</a> </span>.state
|
|
<a href=#1431 id=1431 data-nosnippet>1431</a> .write()
|
|
<a href=#1432 id=1432 data-nosnippet>1432</a> .unwrap()
|
|
<a href=#1433 id=1433 data-nosnippet>1433</a> .dirty_time_windows
|
|
<a href=#1434 id=1434 data-nosnippet>1434</a> .add_dirty_windows(dirty_windows),
|
|
<a href=#1435 id=1435 data-nosnippet>1435</a> }
|
|
<a href=#1436 id=1436 data-nosnippet>1436</a> }
|
|
<a href=#1437 id=1437 data-nosnippet>1437</a>
|
|
<a href=#1438 id=1438 data-nosnippet>1438</a> <span class="doccomment">/// Restore the dirty signal for a plan that was generated but failed before
|
|
<a href=#1439 id=1439 data-nosnippet>1439</a> /// it could prove any checkpoint advancement.
|
|
<a href=#1440 id=1440 data-nosnippet>1440</a> </span><span class="kw">fn </span>restore_dirty_windows_after_failure(<span class="kw-2">&</span><span class="self">self</span>, query: <span class="kw-2">&</span>PlanInfo) {
|
|
<a href=#1441 id=1441 data-nosnippet>1441</a> <span class="self">self</span>.restore_dirty_windows(<span class="kw-2">&</span>query.dirty_restore);
|
|
<a href=#1442 id=1442 data-nosnippet>1442</a> }
|
|
<a href=#1443 id=1443 data-nosnippet>1443</a>
|
|
<a href=#1444 id=1444 data-nosnippet>1444</a> <span class="doccomment">/// Restore scoped windows through `TaskState` so fenced repair can decide
|
|
<a href=#1445 id=1445 data-nosnippet>1445</a> /// whether they go back to pending repair or live dirty state.
|
|
<a href=#1446 id=1446 data-nosnippet>1446</a> </span><span class="kw">fn </span>restore_scoped_dirty_windows(<span class="kw-2">&</span><span class="self">self</span>, filter: <span class="kw-2">&</span>FilterExprInfo) {
|
|
<a href=#1447 id=1447 data-nosnippet>1447</a> <span class="self">self</span>.state.write().unwrap().restore_scoped_windows(filter);
|
|
<a href=#1448 id=1448 data-nosnippet>1448</a> }
|
|
<a href=#1449 id=1449 data-nosnippet>1449</a>
|
|
<a href=#1450 id=1450 data-nosnippet>1450</a> <span class="doccomment">/// Run a fallible scoped operation and restore its consumed windows if plan
|
|
<a href=#1451 id=1451 data-nosnippet>1451</a> /// generation/rewrite fails before execution.
|
|
<a href=#1452 id=1452 data-nosnippet>1452</a> </span><span class="kw">fn </span>restore_scoped_dirty_windows_on_err<T>(
|
|
<a href=#1453 id=1453 data-nosnippet>1453</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1454 id=1454 data-nosnippet>1454</a> filter: <span class="kw-2">&</span>FilterExprInfo,
|
|
<a href=#1455 id=1455 data-nosnippet>1455</a> result: <span class="prelude-ty">Result</span><T, Error>,
|
|
<a href=#1456 id=1456 data-nosnippet>1456</a> ) -> <span class="prelude-ty">Result</span><T, Error> {
|
|
<a href=#1457 id=1457 data-nosnippet>1457</a> result.inspect_err(|<span class="kw">_</span>| {
|
|
<a href=#1458 id=1458 data-nosnippet>1458</a> <span class="self">self</span>.restore_scoped_dirty_windows(filter);
|
|
<a href=#1459 id=1459 data-nosnippet>1459</a> })
|
|
<a href=#1460 id=1460 data-nosnippet>1460</a> }
|
|
<a href=#1461 id=1461 data-nosnippet>1461</a>
|
|
<a href=#1462 id=1462 data-nosnippet>1462</a> <span class="doccomment">/// Restore an unscoped dirty signal consumed by an explicit full-query or
|
|
<a href=#1463 id=1463 data-nosnippet>1463</a> /// incremental-delta plan.
|
|
<a href=#1464 id=1464 data-nosnippet>1464</a> </span><span class="kw">fn </span>restore_unscoped_dirty_windows(<span class="kw-2">&</span><span class="self">self</span>, dirty_windows: <span class="kw-2">&</span>DirtyTimeWindows) {
|
|
<a href=#1465 id=1465 data-nosnippet>1465</a> <span class="self">self</span>.state
|
|
<a href=#1466 id=1466 data-nosnippet>1466</a> .write()
|
|
<a href=#1467 id=1467 data-nosnippet>1467</a> .unwrap()
|
|
<a href=#1468 id=1468 data-nosnippet>1468</a> .dirty_time_windows
|
|
<a href=#1469 id=1469 data-nosnippet>1469</a> .add_dirty_windows(dirty_windows);
|
|
<a href=#1470 id=1470 data-nosnippet>1470</a> }
|
|
<a href=#1471 id=1471 data-nosnippet>1471</a>
|
|
<a href=#1472 id=1472 data-nosnippet>1472</a> <span class="doccomment">/// Run a fallible unscoped operation and restore the dirty signal if it
|
|
<a href=#1473 id=1473 data-nosnippet>1473</a> /// fails before a query is executed.
|
|
<a href=#1474 id=1474 data-nosnippet>1474</a> </span><span class="kw">fn </span>restore_unscoped_dirty_windows_on_err<T>(
|
|
<a href=#1475 id=1475 data-nosnippet>1475</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1476 id=1476 data-nosnippet>1476</a> dirty_windows: <span class="kw-2">&</span>DirtyTimeWindows,
|
|
<a href=#1477 id=1477 data-nosnippet>1477</a> result: <span class="prelude-ty">Result</span><T, Error>,
|
|
<a href=#1478 id=1478 data-nosnippet>1478</a> ) -> <span class="prelude-ty">Result</span><T, Error> {
|
|
<a href=#1479 id=1479 data-nosnippet>1479</a> result.inspect_err(|<span class="kw">_</span>| {
|
|
<a href=#1480 id=1480 data-nosnippet>1480</a> <span class="self">self</span>.restore_unscoped_dirty_windows(dirty_windows);
|
|
<a href=#1481 id=1481 data-nosnippet>1481</a> })
|
|
<a href=#1482 id=1482 data-nosnippet>1482</a> }
|
|
<a href=#1483 id=1483 data-nosnippet>1483</a>
|
|
<a href=#1484 id=1484 data-nosnippet>1484</a> <span class="doccomment">/// Consume the live dirty signal for an unscoped query while keeping a copy
|
|
<a href=#1485 id=1485 data-nosnippet>1485</a> /// that can be restored if planning or execution fails.
|
|
<a href=#1486 id=1486 data-nosnippet>1486</a> </span><span class="kw">fn </span>drain_dirty_windows_signal(<span class="kw-2">&</span><span class="self">self</span>) -> (bool, DirtyTimeWindows) {
|
|
<a href=#1487 id=1487 data-nosnippet>1487</a> <span class="kw">let </span><span class="kw-2">mut </span>state = <span class="self">self</span>.state.write().unwrap();
|
|
<a href=#1488 id=1488 data-nosnippet>1488</a> <span class="kw">let </span>dirty_windows_to_restore = state.dirty_time_windows.clone();
|
|
<a href=#1489 id=1489 data-nosnippet>1489</a> <span class="kw">let </span>is_dirty = !dirty_windows_to_restore.is_empty();
|
|
<a href=#1490 id=1490 data-nosnippet>1490</a> state.dirty_time_windows.clean();
|
|
<a href=#1491 id=1491 data-nosnippet>1491</a> (is_dirty, dirty_windows_to_restore)
|
|
<a href=#1492 id=1492 data-nosnippet>1492</a> }
|
|
<a href=#1493 id=1493 data-nosnippet>1493</a>
|
|
<a href=#1494 id=1494 data-nosnippet>1494</a> <span class="attr">#[allow(clippy::too_many_arguments)]
|
|
<a href=#1495 id=1495 data-nosnippet>1495</a> </span><span class="doccomment">/// Build an unfiltered plan for explicit full-query or incremental-delta
|
|
<a href=#1496 id=1496 data-nosnippet>1496</a> /// coverage. Callers pass the consumed dirty signal for failure restoration.
|
|
<a href=#1497 id=1497 data-nosnippet>1497</a> </span><span class="kw">async fn </span>gen_unfiltered_plan_info(
|
|
<a href=#1498 id=1498 data-nosnippet>1498</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1499 id=1499 data-nosnippet>1499</a> engine: QueryEngineRef,
|
|
<a href=#1500 id=1500 data-nosnippet>1500</a> query_ctx: QueryContextRef,
|
|
<a href=#1501 id=1501 data-nosnippet>1501</a> sink_table_schema: Arc<Schema>,
|
|
<a href=#1502 id=1502 data-nosnippet>1502</a> primary_key_indices: <span class="kw-2">&</span>[usize],
|
|
<a href=#1503 id=1503 data-nosnippet>1503</a> allow_partial: bool,
|
|
<a href=#1504 id=1504 data-nosnippet>1504</a> values: <span class="kw-2">&</span>BTreeMap<String, ScalarValue>,
|
|
<a href=#1505 id=1505 data-nosnippet>1505</a> dirty_windows_to_restore: DirtyTimeWindows,
|
|
<a href=#1506 id=1506 data-nosnippet>1506</a> retention_filter: <span class="prelude-ty">Option</span><(<span class="kw-2">&</span>str, Timestamp, <span class="kw-2">&</span><span class="lifetime">'static </span>str)>,
|
|
<a href=#1507 id=1507 data-nosnippet>1507</a> coverage: QueryCoverage,
|
|
<a href=#1508 id=1508 data-nosnippet>1508</a> ) -> <span class="prelude-ty">Result</span><PlanInfo, Error> {
|
|
<a href=#1509 id=1509 data-nosnippet>1509</a> <span class="kw">let </span><span class="kw-2">mut </span>plan = <span class="self">self</span>.restore_unscoped_dirty_windows_on_err(
|
|
<a href=#1510 id=1510 data-nosnippet>1510</a> <span class="kw-2">&</span>dirty_windows_to_restore,
|
|
<a href=#1511 id=1511 data-nosnippet>1511</a> gen_plan_with_matching_schema_and_values(
|
|
<a href=#1512 id=1512 data-nosnippet>1512</a> <span class="kw-2">&</span><span class="self">self</span>.config.query,
|
|
<a href=#1513 id=1513 data-nosnippet>1513</a> query_ctx,
|
|
<a href=#1514 id=1514 data-nosnippet>1514</a> engine,
|
|
<a href=#1515 id=1515 data-nosnippet>1515</a> sink_table_schema,
|
|
<a href=#1516 id=1516 data-nosnippet>1516</a> primary_key_indices,
|
|
<a href=#1517 id=1517 data-nosnippet>1517</a> allow_partial,
|
|
<a href=#1518 id=1518 data-nosnippet>1518</a> values,
|
|
<a href=#1519 id=1519 data-nosnippet>1519</a> )
|
|
<a href=#1520 id=1520 data-nosnippet>1520</a> .<span class="kw">await</span>,
|
|
<a href=#1521 id=1521 data-nosnippet>1521</a> )<span class="question-mark">?</span>;
|
|
<a href=#1522 id=1522 data-nosnippet>1522</a>
|
|
<a href=#1523 id=1523 data-nosnippet>1523</a> <span class="kw">if let </span><span class="prelude-val">Some</span>((col_name, lower_bound, context)) = retention_filter {
|
|
<a href=#1524 id=1524 data-nosnippet>1524</a> <span class="kw">let </span>lower = <span class="self">self</span>.restore_unscoped_dirty_windows_on_err(
|
|
<a href=#1525 id=1525 data-nosnippet>1525</a> <span class="kw-2">&</span>dirty_windows_to_restore,
|
|
<a href=#1526 id=1526 data-nosnippet>1526</a> to_df_literal(lower_bound),
|
|
<a href=#1527 id=1527 data-nosnippet>1527</a> )<span class="question-mark">?</span>;
|
|
<a href=#1528 id=1528 data-nosnippet>1528</a> <span class="kw">let </span>retention_filter = col(col_name).gt_eq(lit(lower));
|
|
<a href=#1529 id=1529 data-nosnippet>1529</a> <span class="kw">let </span><span class="kw-2">mut </span>add_filter = AddFilterRewriter::new(retention_filter);
|
|
<a href=#1530 id=1530 data-nosnippet>1530</a> plan = <span class="self">self</span>.restore_unscoped_dirty_windows_on_err(
|
|
<a href=#1531 id=1531 data-nosnippet>1531</a> <span class="kw-2">&</span>dirty_windows_to_restore,
|
|
<a href=#1532 id=1532 data-nosnippet>1532</a> plan.clone()
|
|
<a href=#1533 id=1533 data-nosnippet>1533</a> .rewrite(<span class="kw-2">&mut </span>add_filter)
|
|
<a href=#1534 id=1534 data-nosnippet>1534</a> .with_context(|<span class="kw">_</span>| DatafusionSnafu {
|
|
<a href=#1535 id=1535 data-nosnippet>1535</a> context: <span class="macro">format!</span>(
|
|
<a href=#1536 id=1536 data-nosnippet>1536</a> <span class="string">"Failed to apply {context} expire_after filter to plan:\n {}\n"</span>,
|
|
<a href=#1537 id=1537 data-nosnippet>1537</a> plan
|
|
<a href=#1538 id=1538 data-nosnippet>1538</a> ),
|
|
<a href=#1539 id=1539 data-nosnippet>1539</a> })
|
|
<a href=#1540 id=1540 data-nosnippet>1540</a> .map(|rewrite| rewrite.data),
|
|
<a href=#1541 id=1541 data-nosnippet>1541</a> )<span class="question-mark">?</span>;
|
|
<a href=#1542 id=1542 data-nosnippet>1542</a> }
|
|
<a href=#1543 id=1543 data-nosnippet>1543</a>
|
|
<a href=#1544 id=1544 data-nosnippet>1544</a> <span class="prelude-val">Ok</span>(PlanInfo {
|
|
<a href=#1545 id=1545 data-nosnippet>1545</a> plan,
|
|
<a href=#1546 id=1546 data-nosnippet>1546</a> dirty_restore: DirtyRestore::Unscoped(dirty_windows_to_restore),
|
|
<a href=#1547 id=1547 data-nosnippet>1547</a> coverage,
|
|
<a href=#1548 id=1548 data-nosnippet>1548</a> })
|
|
<a href=#1549 id=1549 data-nosnippet>1549</a> }
|
|
<a href=#1550 id=1550 data-nosnippet>1550</a>
|
|
<a href=#1551 id=1551 data-nosnippet>1551</a> <span class="attr">#[allow(clippy::too_many_arguments)]
|
|
<a href=#1552 id=1552 data-nosnippet>1552</a> </span><span class="doccomment">/// Build an unfiltered plan only when the live dirty signal was present;
|
|
<a href=#1553 id=1553 data-nosnippet>1553</a> /// otherwise skip this round without querying.
|
|
<a href=#1554 id=1554 data-nosnippet>1554</a> </span><span class="kw">async fn </span>gen_unfiltered_plan_info_if_dirty(
|
|
<a href=#1555 id=1555 data-nosnippet>1555</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1556 id=1556 data-nosnippet>1556</a> engine: QueryEngineRef,
|
|
<a href=#1557 id=1557 data-nosnippet>1557</a> query_ctx: QueryContextRef,
|
|
<a href=#1558 id=1558 data-nosnippet>1558</a> sink_table_schema: Arc<Schema>,
|
|
<a href=#1559 id=1559 data-nosnippet>1559</a> primary_key_indices: <span class="kw-2">&</span>[usize],
|
|
<a href=#1560 id=1560 data-nosnippet>1560</a> allow_partial: bool,
|
|
<a href=#1561 id=1561 data-nosnippet>1561</a> values: <span class="kw-2">&</span>BTreeMap<String, ScalarValue>,
|
|
<a href=#1562 id=1562 data-nosnippet>1562</a> retention_filter: <span class="prelude-ty">Option</span><(<span class="kw-2">&</span>str, Timestamp, <span class="kw-2">&</span><span class="lifetime">'static </span>str)>,
|
|
<a href=#1563 id=1563 data-nosnippet>1563</a> coverage: QueryCoverage,
|
|
<a href=#1564 id=1564 data-nosnippet>1564</a> ) -> <span class="prelude-ty">Result</span><<span class="prelude-ty">Option</span><PlanInfo>, Error> {
|
|
<a href=#1565 id=1565 data-nosnippet>1565</a> <span class="kw">let </span>(is_dirty, dirty_windows_to_restore) = <span class="self">self</span>.drain_dirty_windows_signal();
|
|
<a href=#1566 id=1566 data-nosnippet>1566</a> <span class="kw">if </span>!is_dirty {
|
|
<a href=#1567 id=1567 data-nosnippet>1567</a> <span class="macro">debug!</span>(<span class="string">"Flow id={:?}, no new data, not update"</span>, <span class="self">self</span>.config.flow_id);
|
|
<a href=#1568 id=1568 data-nosnippet>1568</a> <span class="kw">return </span><span class="prelude-val">Ok</span>(<span class="prelude-val">None</span>);
|
|
<a href=#1569 id=1569 data-nosnippet>1569</a> }
|
|
<a href=#1570 id=1570 data-nosnippet>1570</a>
|
|
<a href=#1571 id=1571 data-nosnippet>1571</a> <span class="self">self</span>.gen_unfiltered_plan_info(
|
|
<a href=#1572 id=1572 data-nosnippet>1572</a> engine,
|
|
<a href=#1573 id=1573 data-nosnippet>1573</a> query_ctx,
|
|
<a href=#1574 id=1574 data-nosnippet>1574</a> sink_table_schema,
|
|
<a href=#1575 id=1575 data-nosnippet>1575</a> primary_key_indices,
|
|
<a href=#1576 id=1576 data-nosnippet>1576</a> allow_partial,
|
|
<a href=#1577 id=1577 data-nosnippet>1577</a> values,
|
|
<a href=#1578 id=1578 data-nosnippet>1578</a> dirty_windows_to_restore,
|
|
<a href=#1579 id=1579 data-nosnippet>1579</a> retention_filter,
|
|
<a href=#1580 id=1580 data-nosnippet>1580</a> coverage,
|
|
<a href=#1581 id=1581 data-nosnippet>1581</a> )
|
|
<a href=#1582 id=1582 data-nosnippet>1582</a> .<span class="kw">await
|
|
<a href=#1583 id=1583 data-nosnippet>1583</a> </span>.map(<span class="prelude-val">Some</span>)
|
|
<a href=#1584 id=1584 data-nosnippet>1584</a> }
|
|
<a href=#1585 id=1585 data-nosnippet>1585</a>
|
|
<a href=#1586 id=1586 data-nosnippet>1586</a> <span class="kw">fn </span>handle_executed_query_failure(<span class="kw-2">&</span><span class="self">self</span>, query: <span class="prelude-ty">Option</span><<span class="kw-2">&</span>PlanInfo>) {
|
|
<a href=#1587 id=1587 data-nosnippet>1587</a> <span class="kw">if let </span><span class="prelude-val">Some</span>(query) = query {
|
|
<a href=#1588 id=1588 data-nosnippet>1588</a> <span class="self">self</span>.restore_dirty_windows_after_failure(query);
|
|
<a href=#1589 id=1589 data-nosnippet>1589</a> }
|
|
<a href=#1590 id=1590 data-nosnippet>1590</a> }
|
|
<a href=#1591 id=1591 data-nosnippet>1591</a>
|
|
<a href=#1592 id=1592 data-nosnippet>1592</a> <span class="doccomment">/// start executing query in a loop, break when receive shutdown signal
|
|
<a href=#1593 id=1593 data-nosnippet>1593</a> ///
|
|
<a href=#1594 id=1594 data-nosnippet>1594</a> /// any error will be logged when executing query.
|
|
<a href=#1595 id=1595 data-nosnippet>1595</a> ///
|
|
<a href=#1596 id=1596 data-nosnippet>1596</a> /// Dispatches to:
|
|
<a href=#1597 id=1597 data-nosnippet>1597</a> /// - scheduled loop when `flow_eval_interval.is_some()`
|
|
<a href=#1598 id=1598 data-nosnippet>1598</a> /// - adaptive dirty-window loop otherwise
|
|
<a href=#1599 id=1599 data-nosnippet>1599</a> </span><span class="kw">pub async fn </span>start_executing_loop(
|
|
<a href=#1600 id=1600 data-nosnippet>1600</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1601 id=1601 data-nosnippet>1601</a> engine: QueryEngineRef,
|
|
<a href=#1602 id=1602 data-nosnippet>1602</a> frontend_client: Arc<FrontendClient>,
|
|
<a href=#1603 id=1603 data-nosnippet>1603</a> ) {
|
|
<a href=#1604 id=1604 data-nosnippet>1604</a> <span class="kw">if </span><span class="self">self</span>.config.flow_eval_interval.is_some() {
|
|
<a href=#1605 id=1605 data-nosnippet>1605</a> <span class="self">self</span>.start_scheduled_loop(engine, frontend_client).<span class="kw">await</span>;
|
|
<a href=#1606 id=1606 data-nosnippet>1606</a> } <span class="kw">else </span>{
|
|
<a href=#1607 id=1607 data-nosnippet>1607</a> <span class="self">self</span>.start_adaptive_loop(engine, frontend_client).<span class="kw">await</span>;
|
|
<a href=#1608 id=1608 data-nosnippet>1608</a> }
|
|
<a href=#1609 id=1609 data-nosnippet>1609</a> }
|
|
<a href=#1610 id=1610 data-nosnippet>1610</a>
|
|
<a href=#1611 id=1611 data-nosnippet>1611</a> <span class="doccomment">/// Scheduled batching loop for flows with `EVAL INTERVAL`.
|
|
<a href=#1612 id=1612 data-nosnippet>1612</a> ///
|
|
<a href=#1613 id=1613 data-nosnippet>1613</a> /// Uses the pre-parsed `EvalSchedule` from `TaskConfig` and selects due
|
|
<a href=#1614 id=1614 data-nosnippet>1614</a> /// scheduled times using bounded catch-up semantics. Each scheduled time is the
|
|
<a href=#1615 id=1615 data-nosnippet>1615</a> /// scheduled evaluation time used as logical `now()` for that attempt.
|
|
<a href=#1616 id=1616 data-nosnippet>1616</a> /// Each attempt temporarily sets `flow.scheduled_time_millis` on the
|
|
<a href=#1617 id=1617 data-nosnippet>1617</a> /// task's `QueryContext` and executes under the existing `execution_lock`.
|
|
<a href=#1618 id=1618 data-nosnippet>1618</a> /// After every attempt (success, no-op, or failure) the in-memory
|
|
<a href=#1619 id=1619 data-nosnippet>1619</a> /// cursor advances.
|
|
<a href=#1620 id=1620 data-nosnippet>1620</a> </span><span class="kw">async fn </span>start_scheduled_loop(
|
|
<a href=#1621 id=1621 data-nosnippet>1621</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1622 id=1622 data-nosnippet>1622</a> engine: QueryEngineRef,
|
|
<a href=#1623 id=1623 data-nosnippet>1623</a> frontend_client: Arc<FrontendClient>,
|
|
<a href=#1624 id=1624 data-nosnippet>1624</a> ) {
|
|
<a href=#1625 id=1625 data-nosnippet>1625</a> <span class="kw">let </span>flow_id_str = <span class="self">self</span>.config.flow_id.to_string();
|
|
<a href=#1626 id=1626 data-nosnippet>1626</a>
|
|
<a href=#1627 id=1627 data-nosnippet>1627</a> <span class="kw">let </span>schedule = <span class="kw">match </span><span class="kw-2">&</span><span class="self">self</span>.config.eval_schedule {
|
|
<a href=#1628 id=1628 data-nosnippet>1628</a> <span class="prelude-val">Some</span>(s) => s.clone(),
|
|
<a href=#1629 id=1629 data-nosnippet>1629</a> <span class="prelude-val">None </span>=> {
|
|
<a href=#1630 id=1630 data-nosnippet>1630</a> <span class="kw">let </span>eval_interval_secs = <span class="self">self
|
|
<a href=#1631 id=1631 data-nosnippet>1631</a> </span>.config
|
|
<a href=#1632 id=1632 data-nosnippet>1632</a> .flow_eval_interval
|
|
<a href=#1633 id=1633 data-nosnippet>1633</a> .map(|d| d.as_secs() <span class="kw">as </span>i64)
|
|
<a href=#1634 id=1634 data-nosnippet>1634</a> .expect(<span class="string">"checked by caller"</span>);
|
|
<a href=#1635 id=1635 data-nosnippet>1635</a>
|
|
<a href=#1636 id=1636 data-nosnippet>1636</a> <span class="comment">// Fallback: no typed config provided. Compute defaults
|
|
<a href=#1637 id=1637 data-nosnippet>1637</a> // anchored at epoch/start=0.
|
|
<a href=#1638 id=1638 data-nosnippet>1638</a> </span><span class="kw">match </span>EvalSchedule::from_config(<span class="prelude-val">Some</span>(eval_interval_secs), <span class="prelude-val">None</span>) {
|
|
<a href=#1639 id=1639 data-nosnippet>1639</a> <span class="prelude-val">Ok</span>(<span class="prelude-val">Some</span>(s)) => s,
|
|
<a href=#1640 id=1640 data-nosnippet>1640</a> <span class="prelude-val">Ok</span>(<span class="prelude-val">None</span>) => {
|
|
<a href=#1641 id=1641 data-nosnippet>1641</a> <span class="macro">warn!</span>(
|
|
<a href=#1642 id=1642 data-nosnippet>1642</a> <span class="string">"Flow {}: EVAL INTERVAL set but no schedule parsed; exiting loop"</span>,
|
|
<a href=#1643 id=1643 data-nosnippet>1643</a> flow_id_str
|
|
<a href=#1644 id=1644 data-nosnippet>1644</a> );
|
|
<a href=#1645 id=1645 data-nosnippet>1645</a> <span class="kw">return</span>;
|
|
<a href=#1646 id=1646 data-nosnippet>1646</a> }
|
|
<a href=#1647 id=1647 data-nosnippet>1647</a> <span class="prelude-val">Err</span>(e) => {
|
|
<a href=#1648 id=1648 data-nosnippet>1648</a> <span class="macro">warn!</span>(
|
|
<a href=#1649 id=1649 data-nosnippet>1649</a> <span class="string">"Flow {}: Failed to parse eval schedule: {}; exiting loop"</span>,
|
|
<a href=#1650 id=1650 data-nosnippet>1650</a> flow_id_str, e
|
|
<a href=#1651 id=1651 data-nosnippet>1651</a> );
|
|
<a href=#1652 id=1652 data-nosnippet>1652</a> <span class="kw">return</span>;
|
|
<a href=#1653 id=1653 data-nosnippet>1653</a> }
|
|
<a href=#1654 id=1654 data-nosnippet>1654</a> }
|
|
<a href=#1655 id=1655 data-nosnippet>1655</a> }
|
|
<a href=#1656 id=1656 data-nosnippet>1656</a> };
|
|
<a href=#1657 id=1657 data-nosnippet>1657</a>
|
|
<a href=#1658 id=1658 data-nosnippet>1658</a> <span class="comment">// Initial cursor is one interval before start so the first due
|
|
<a href=#1659 id=1659 data-nosnippet>1659</a> // scheduled time is `start_secs`. An unrepresentable difference is an
|
|
<a href=#1660 id=1660 data-nosnippet>1660</a> // explicit error, never a saturated cursor that would silently skip
|
|
<a href=#1661 id=1661 data-nosnippet>1661</a> // the first scheduled tick.
|
|
<a href=#1662 id=1662 data-nosnippet>1662</a> </span><span class="kw">let </span><span class="kw-2">mut </span>cursor_secs =
|
|
<a href=#1663 id=1663 data-nosnippet>1663</a> <span class="kw">match </span>initial_schedule_cursor(schedule.start_secs, schedule.interval_secs) {
|
|
<a href=#1664 id=1664 data-nosnippet>1664</a> <span class="prelude-val">Ok</span>(cursor) => cursor,
|
|
<a href=#1665 id=1665 data-nosnippet>1665</a> <span class="prelude-val">Err</span>(e) => {
|
|
<a href=#1666 id=1666 data-nosnippet>1666</a> <span class="macro">warn!</span>(
|
|
<a href=#1667 id=1667 data-nosnippet>1667</a> <span class="string">"Flow {}: invalid eval schedule, exiting loop: {}"</span>,
|
|
<a href=#1668 id=1668 data-nosnippet>1668</a> flow_id_str, e
|
|
<a href=#1669 id=1669 data-nosnippet>1669</a> );
|
|
<a href=#1670 id=1670 data-nosnippet>1670</a> <span class="kw">return</span>;
|
|
<a href=#1671 id=1671 data-nosnippet>1671</a> }
|
|
<a href=#1672 id=1672 data-nosnippet>1672</a> };
|
|
<a href=#1673 id=1673 data-nosnippet>1673</a>
|
|
<a href=#1674 id=1674 data-nosnippet>1674</a> <span class="macro">info!</span>(
|
|
<a href=#1675 id=1675 data-nosnippet>1675</a> <span class="string">"Flow {}: entering scheduled loop, interval={}s, start={}, anchor={}, policy={:?}, max_runs={}, max_lag={}s"</span>,
|
|
<a href=#1676 id=1676 data-nosnippet>1676</a> flow_id_str,
|
|
<a href=#1677 id=1677 data-nosnippet>1677</a> schedule.interval_secs,
|
|
<a href=#1678 id=1678 data-nosnippet>1678</a> schedule.start_secs,
|
|
<a href=#1679 id=1679 data-nosnippet>1679</a> schedule.anchor_secs,
|
|
<a href=#1680 id=1680 data-nosnippet>1680</a> schedule.missed_tick_policy,
|
|
<a href=#1681 id=1681 data-nosnippet>1681</a> schedule.max_runs,
|
|
<a href=#1682 id=1682 data-nosnippet>1682</a> schedule.max_lag_secs,
|
|
<a href=#1683 id=1683 data-nosnippet>1683</a> );
|
|
<a href=#1684 id=1684 data-nosnippet>1684</a>
|
|
<a href=#1685 id=1685 data-nosnippet>1685</a> <span class="kw">loop </span>{
|
|
<a href=#1686 id=1686 data-nosnippet>1686</a> <span class="kw">if </span><span class="self">self</span>.is_shutdown_signaled() {
|
|
<a href=#1687 id=1687 data-nosnippet>1687</a> <span class="kw">break</span>;
|
|
<a href=#1688 id=1688 data-nosnippet>1688</a> }
|
|
<a href=#1689 id=1689 data-nosnippet>1689</a>
|
|
<a href=#1690 id=1690 data-nosnippet>1690</a> <span class="kw">let </span>wall_now_secs = wall_clock_unix_secs();
|
|
<a href=#1691 id=1691 data-nosnippet>1691</a>
|
|
<a href=#1692 id=1692 data-nosnippet>1692</a> <span class="kw">let </span>due = <span class="kw">match </span>select_due_scheduled_times(<span class="kw-2">&</span>schedule, cursor_secs, wall_now_secs) {
|
|
<a href=#1693 id=1693 data-nosnippet>1693</a> <span class="prelude-val">Ok</span>(d) => d,
|
|
<a href=#1694 id=1694 data-nosnippet>1694</a> <span class="prelude-val">Err</span>(e) => {
|
|
<a href=#1695 id=1695 data-nosnippet>1695</a> <span class="macro">warn!</span>(
|
|
<a href=#1696 id=1696 data-nosnippet>1696</a> <span class="string">"Flow {}: invalid eval schedule, exiting loop: {}"</span>,
|
|
<a href=#1697 id=1697 data-nosnippet>1697</a> flow_id_str, e
|
|
<a href=#1698 id=1698 data-nosnippet>1698</a> );
|
|
<a href=#1699 id=1699 data-nosnippet>1699</a> <span class="kw">return</span>;
|
|
<a href=#1700 id=1700 data-nosnippet>1700</a> }
|
|
<a href=#1701 id=1701 data-nosnippet>1701</a> };
|
|
<a href=#1702 id=1702 data-nosnippet>1702</a>
|
|
<a href=#1703 id=1703 data-nosnippet>1703</a> <span class="kw">if </span>due.scheduled_times_secs.is_empty() {
|
|
<a href=#1704 id=1704 data-nosnippet>1704</a> <span class="kw">if </span>due.skipped > <span class="number">0 </span>{
|
|
<a href=#1705 id=1705 data-nosnippet>1705</a> <span class="macro">warn!</span>(
|
|
<a href=#1706 id=1706 data-nosnippet>1706</a> <span class="string">"Flow {}: all {} due scheduled times skipped by max-lag, advancing cursor to wall-clock ({wall_now_secs}) to avoid re-skipping"</span>,
|
|
<a href=#1707 id=1707 data-nosnippet>1707</a> flow_id_str, due.skipped
|
|
<a href=#1708 id=1708 data-nosnippet>1708</a> );
|
|
<a href=#1709 id=1709 data-nosnippet>1709</a> cursor_secs = wall_now_secs;
|
|
<a href=#1710 id=1710 data-nosnippet>1710</a> <span class="kw">continue</span>;
|
|
<a href=#1711 id=1711 data-nosnippet>1711</a> }
|
|
<a href=#1712 id=1712 data-nosnippet>1712</a>
|
|
<a href=#1713 id=1713 data-nosnippet>1713</a> <span class="comment">// No due yet — sleep until the next scheduled time.
|
|
<a href=#1714 id=1714 data-nosnippet>1714</a> </span><span class="kw">let </span>next = <span class="kw">match </span>schedule.next_scheduled_time_after(cursor_secs) {
|
|
<a href=#1715 id=1715 data-nosnippet>1715</a> <span class="prelude-val">Ok</span>(next) => next,
|
|
<a href=#1716 id=1716 data-nosnippet>1716</a> <span class="prelude-val">Err</span>(e) => {
|
|
<a href=#1717 id=1717 data-nosnippet>1717</a> <span class="macro">warn!</span>(
|
|
<a href=#1718 id=1718 data-nosnippet>1718</a> <span class="string">"Flow {}: cannot advance eval schedule past cursor {cursor_secs}: {e}; exiting loop"</span>,
|
|
<a href=#1719 id=1719 data-nosnippet>1719</a> flow_id_str
|
|
<a href=#1720 id=1720 data-nosnippet>1720</a> );
|
|
<a href=#1721 id=1721 data-nosnippet>1721</a> <span class="kw">return</span>;
|
|
<a href=#1722 id=1722 data-nosnippet>1722</a> }
|
|
<a href=#1723 id=1723 data-nosnippet>1723</a> };
|
|
<a href=#1724 id=1724 data-nosnippet>1724</a> <span class="kw">if </span>next <= wall_now_secs {
|
|
<a href=#1725 id=1725 data-nosnippet>1725</a> <span class="comment">// Shouldn't happen given select_due_scheduled_times returned empty,
|
|
<a href=#1726 id=1726 data-nosnippet>1726</a> // but guard against clock skew / logic error.
|
|
<a href=#1727 id=1727 data-nosnippet>1727</a> </span>cursor_secs = wall_now_secs;
|
|
<a href=#1728 id=1728 data-nosnippet>1728</a> <span class="kw">continue</span>;
|
|
<a href=#1729 id=1729 data-nosnippet>1729</a> }
|
|
<a href=#1730 id=1730 data-nosnippet>1730</a> <span class="kw">let </span>wait_secs = <span class="kw">match </span>sleep_delta_secs(next, wall_now_secs) {
|
|
<a href=#1731 id=1731 data-nosnippet>1731</a> <span class="prelude-val">Ok</span>(wait_secs) => wait_secs,
|
|
<a href=#1732 id=1732 data-nosnippet>1732</a> <span class="prelude-val">Err</span>(e) => {
|
|
<a href=#1733 id=1733 data-nosnippet>1733</a> <span class="macro">warn!</span>(
|
|
<a href=#1734 id=1734 data-nosnippet>1734</a> <span class="string">"Flow {}: cannot sleep until next scheduled time {}: {e}; exiting loop"</span>,
|
|
<a href=#1735 id=1735 data-nosnippet>1735</a> flow_id_str, next
|
|
<a href=#1736 id=1736 data-nosnippet>1736</a> );
|
|
<a href=#1737 id=1737 data-nosnippet>1737</a> <span class="kw">return</span>;
|
|
<a href=#1738 id=1738 data-nosnippet>1738</a> }
|
|
<a href=#1739 id=1739 data-nosnippet>1739</a> };
|
|
<a href=#1740 id=1740 data-nosnippet>1740</a> <span class="kw">let </span>wait_dur = Duration::from_secs(wait_secs);
|
|
<a href=#1741 id=1741 data-nosnippet>1741</a> <span class="macro">debug!</span>(
|
|
<a href=#1742 id=1742 data-nosnippet>1742</a> <span class="string">"Flow {}: no due scheduled times, sleeping for {}s until next scheduled time at {}"</span>,
|
|
<a href=#1743 id=1743 data-nosnippet>1743</a> flow_id_str, wait_secs, next
|
|
<a href=#1744 id=1744 data-nosnippet>1744</a> );
|
|
<a href=#1745 id=1745 data-nosnippet>1745</a> tokio::time::sleep(wait_dur).<span class="kw">await</span>;
|
|
<a href=#1746 id=1746 data-nosnippet>1746</a> <span class="kw">continue</span>;
|
|
<a href=#1747 id=1747 data-nosnippet>1747</a> }
|
|
<a href=#1748 id=1748 data-nosnippet>1748</a>
|
|
<a href=#1749 id=1749 data-nosnippet>1749</a> <span class="kw">if </span>due.skipped > <span class="number">0 </span>{
|
|
<a href=#1750 id=1750 data-nosnippet>1750</a> <span class="macro">info!</span>(
|
|
<a href=#1751 id=1751 data-nosnippet>1751</a> <span class="string">"Flow {}: {} due scheduled times, {} skipped (catch-up)"</span>,
|
|
<a href=#1752 id=1752 data-nosnippet>1752</a> flow_id_str,
|
|
<a href=#1753 id=1753 data-nosnippet>1753</a> due.scheduled_times_secs.len(),
|
|
<a href=#1754 id=1754 data-nosnippet>1754</a> due.skipped
|
|
<a href=#1755 id=1755 data-nosnippet>1755</a> );
|
|
<a href=#1756 id=1756 data-nosnippet>1756</a> }
|
|
<a href=#1757 id=1757 data-nosnippet>1757</a>
|
|
<a href=#1758 id=1758 data-nosnippet>1758</a> <span class="comment">// Execute scheduled times oldest → newest.
|
|
<a href=#1759 id=1759 data-nosnippet>1759</a> </span><span class="kw">for </span>scheduled_time_secs <span class="kw">in </span><span class="kw-2">&</span>due.scheduled_times_secs {
|
|
<a href=#1760 id=1760 data-nosnippet>1760</a> <span class="kw">if </span><span class="self">self</span>.is_shutdown_signaled() {
|
|
<a href=#1761 id=1761 data-nosnippet>1761</a> <span class="kw">break</span>;
|
|
<a href=#1762 id=1762 data-nosnippet>1762</a> }
|
|
<a href=#1763 id=1763 data-nosnippet>1763</a>
|
|
<a href=#1764 id=1764 data-nosnippet>1764</a> METRIC_FLOW_BATCHING_ENGINE_START_QUERY_CNT
|
|
<a href=#1765 id=1765 data-nosnippet>1765</a> .with_label_values(<span class="kw-2">&</span>[<span class="kw-2">&</span>flow_id_str])
|
|
<a href=#1766 id=1766 data-nosnippet>1766</a> .inc();
|
|
<a href=#1767 id=1767 data-nosnippet>1767</a>
|
|
<a href=#1768 id=1768 data-nosnippet>1768</a> <span class="kw">let </span>outcome = <span class="self">self
|
|
<a href=#1769 id=1769 data-nosnippet>1769</a> </span>.execute_once_serialized_at_scheduled_time(
|
|
<a href=#1770 id=1770 data-nosnippet>1770</a> <span class="kw-2">&</span>engine,
|
|
<a href=#1771 id=1771 data-nosnippet>1771</a> <span class="kw-2">&</span>frontend_client,
|
|
<a href=#1772 id=1772 data-nosnippet>1772</a> <span class="kw-2">*</span>scheduled_time_secs,
|
|
<a href=#1773 id=1773 data-nosnippet>1773</a> )
|
|
<a href=#1774 id=1774 data-nosnippet>1774</a> .<span class="kw">await</span>;
|
|
<a href=#1775 id=1775 data-nosnippet>1775</a>
|
|
<a href=#1776 id=1776 data-nosnippet>1776</a> <span class="comment">// Advance cursor regardless of outcome.
|
|
<a href=#1777 id=1777 data-nosnippet>1777</a> </span>cursor_secs = <span class="kw-2">*</span>scheduled_time_secs;
|
|
<a href=#1778 id=1778 data-nosnippet>1778</a>
|
|
<a href=#1779 id=1779 data-nosnippet>1779</a> <span class="kw">match </span>outcome.result {
|
|
<a href=#1780 id=1780 data-nosnippet>1780</a> <span class="prelude-val">Ok</span>(<span class="prelude-val">Some</span>((rows, elapsed))) => {
|
|
<a href=#1781 id=1781 data-nosnippet>1781</a> <span class="macro">debug!</span>(
|
|
<a href=#1782 id=1782 data-nosnippet>1782</a> <span class="string">"Flow {}: scheduled time {} completed, rows={}, elapsed={:?}"</span>,
|
|
<a href=#1783 id=1783 data-nosnippet>1783</a> flow_id_str, scheduled_time_secs, rows, elapsed
|
|
<a href=#1784 id=1784 data-nosnippet>1784</a> );
|
|
<a href=#1785 id=1785 data-nosnippet>1785</a> }
|
|
<a href=#1786 id=1786 data-nosnippet>1786</a> <span class="prelude-val">Ok</span>(<span class="prelude-val">None</span>) => {
|
|
<a href=#1787 id=1787 data-nosnippet>1787</a> <span class="macro">debug!</span>(
|
|
<a href=#1788 id=1788 data-nosnippet>1788</a> <span class="string">"Flow {}: scheduled time {} produced no query (no dirty signal or no-op)"</span>,
|
|
<a href=#1789 id=1789 data-nosnippet>1789</a> flow_id_str, scheduled_time_secs
|
|
<a href=#1790 id=1790 data-nosnippet>1790</a> );
|
|
<a href=#1791 id=1791 data-nosnippet>1791</a> }
|
|
<a href=#1792 id=1792 data-nosnippet>1792</a> <span class="prelude-val">Err</span>(err) => {
|
|
<a href=#1793 id=1793 data-nosnippet>1793</a> <span class="macro">warn!</span>(
|
|
<a href=#1794 id=1794 data-nosnippet>1794</a> <span class="string">"Flow {}: scheduled time {} failed: {:?}"</span>,
|
|
<a href=#1795 id=1795 data-nosnippet>1795</a> flow_id_str, scheduled_time_secs, err
|
|
<a href=#1796 id=1796 data-nosnippet>1796</a> );
|
|
<a href=#1797 id=1797 data-nosnippet>1797</a> METRIC_FLOW_BATCHING_ENGINE_ERROR_CNT
|
|
<a href=#1798 id=1798 data-nosnippet>1798</a> .with_label_values(<span class="kw-2">&</span>[<span class="kw-2">&</span>flow_id_str])
|
|
<a href=#1799 id=1799 data-nosnippet>1799</a> .inc();
|
|
<a href=#1800 id=1800 data-nosnippet>1800</a> <span class="comment">// Dirty-window restoration is handled by the
|
|
<a href=#1801 id=1801 data-nosnippet>1801</a> // existing `handle_executed_query_failure` inside
|
|
<a href=#1802 id=1802 data-nosnippet>1802</a> // default execution.
|
|
<a href=#1803 id=1803 data-nosnippet>1803</a> </span>}
|
|
<a href=#1804 id=1804 data-nosnippet>1804</a> }
|
|
<a href=#1805 id=1805 data-nosnippet>1805</a> }
|
|
<a href=#1806 id=1806 data-nosnippet>1806</a> }
|
|
<a href=#1807 id=1807 data-nosnippet>1807</a> }
|
|
<a href=#1808 id=1808 data-nosnippet>1808</a>
|
|
<a href=#1809 id=1809 data-nosnippet>1809</a> <span class="doccomment">/// Existing adaptive dirty-window loop for flows without `EVAL INTERVAL`.
|
|
<a href=#1810 id=1810 data-nosnippet>1810</a> </span><span class="kw">async fn </span>start_adaptive_loop(
|
|
<a href=#1811 id=1811 data-nosnippet>1811</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1812 id=1812 data-nosnippet>1812</a> engine: QueryEngineRef,
|
|
<a href=#1813 id=1813 data-nosnippet>1813</a> frontend_client: Arc<FrontendClient>,
|
|
<a href=#1814 id=1814 data-nosnippet>1814</a> ) {
|
|
<a href=#1815 id=1815 data-nosnippet>1815</a> <span class="kw">let </span>flow_id_str = <span class="self">self</span>.config.flow_id.to_string();
|
|
<a href=#1816 id=1816 data-nosnippet>1816</a> <span class="kw">let </span><span class="kw-2">mut </span>max_window_cnt = <span class="prelude-val">None</span>;
|
|
<a href=#1817 id=1817 data-nosnippet>1817</a> <span class="kw">loop </span>{
|
|
<a href=#1818 id=1818 data-nosnippet>1818</a> <span class="kw">if </span><span class="self">self</span>.is_shutdown_signaled() {
|
|
<a href=#1819 id=1819 data-nosnippet>1819</a> <span class="kw">break</span>;
|
|
<a href=#1820 id=1820 data-nosnippet>1820</a> }
|
|
<a href=#1821 id=1821 data-nosnippet>1821</a> METRIC_FLOW_BATCHING_ENGINE_START_QUERY_CNT
|
|
<a href=#1822 id=1822 data-nosnippet>1822</a> .with_label_values(<span class="kw-2">&</span>[<span class="kw-2">&</span>flow_id_str])
|
|
<a href=#1823 id=1823 data-nosnippet>1823</a> .inc();
|
|
<a href=#1824 id=1824 data-nosnippet>1824</a>
|
|
<a href=#1825 id=1825 data-nosnippet>1825</a> <span class="kw">let </span>min_refresh = <span class="self">self</span>.config.batch_opts.experimental_min_refresh_duration;
|
|
<a href=#1826 id=1826 data-nosnippet>1826</a>
|
|
<a href=#1827 id=1827 data-nosnippet>1827</a> <span class="kw">let </span>outcome = <span class="self">self
|
|
<a href=#1828 id=1828 data-nosnippet>1828</a> </span>.execute_once_serialized_with_outcome(<span class="kw-2">&</span>engine, <span class="kw-2">&</span>frontend_client, max_window_cnt)
|
|
<a href=#1829 id=1829 data-nosnippet>1829</a> .<span class="kw">await</span>;
|
|
<a href=#1830 id=1830 data-nosnippet>1830</a>
|
|
<a href=#1831 id=1831 data-nosnippet>1831</a> <span class="kw">match </span>outcome.result {
|
|
<a href=#1832 id=1832 data-nosnippet>1832</a> <span class="prelude-val">Ok</span>(<span class="prelude-val">Some</span>(<span class="kw">_</span>)) => {
|
|
<a href=#1833 id=1833 data-nosnippet>1833</a> max_window_cnt = max_window_cnt.map(|cnt| {
|
|
<a href=#1834 id=1834 data-nosnippet>1834</a> (cnt + <span class="number">1</span>).min(<span class="self">self</span>.config.batch_opts.experimental_max_filter_num_per_query)
|
|
<a href=#1835 id=1835 data-nosnippet>1835</a> });
|
|
<a href=#1836 id=1836 data-nosnippet>1836</a>
|
|
<a href=#1837 id=1837 data-nosnippet>1837</a> <span class="kw">let </span>sleep_until = {
|
|
<a href=#1838 id=1838 data-nosnippet>1838</a> <span class="kw">let </span>state = <span class="self">self</span>.state.write().unwrap();
|
|
<a href=#1839 id=1839 data-nosnippet>1839</a>
|
|
<a href=#1840 id=1840 data-nosnippet>1840</a> <span class="kw">let </span>time_window_size = <span class="self">self
|
|
<a href=#1841 id=1841 data-nosnippet>1841</a> </span>.config
|
|
<a href=#1842 id=1842 data-nosnippet>1842</a> .time_window_expr
|
|
<a href=#1843 id=1843 data-nosnippet>1843</a> .as_ref()
|
|
<a href=#1844 id=1844 data-nosnippet>1844</a> .and_then(|t| <span class="kw-2">*</span>t.time_window_size());
|
|
<a href=#1845 id=1845 data-nosnippet>1845</a>
|
|
<a href=#1846 id=1846 data-nosnippet>1846</a> <span class="kw">let </span>prefer_short_incremental_cadence = state.checkpoint_mode()
|
|
<a href=#1847 id=1847 data-nosnippet>1847</a> == CheckpointMode::Incremental
|
|
<a href=#1848 id=1848 data-nosnippet>1848</a> && !state.is_incremental_disabled();
|
|
<a href=#1849 id=1849 data-nosnippet>1849</a>
|
|
<a href=#1850 id=1850 data-nosnippet>1850</a> state.get_next_start_query_time(
|
|
<a href=#1851 id=1851 data-nosnippet>1851</a> <span class="self">self</span>.config.flow_id,
|
|
<a href=#1852 id=1852 data-nosnippet>1852</a> <span class="kw-2">&</span>time_window_size,
|
|
<a href=#1853 id=1853 data-nosnippet>1853</a> min_refresh,
|
|
<a href=#1854 id=1854 data-nosnippet>1854</a> <span class="prelude-val">Some</span>(<span class="self">self</span>.config.batch_opts.query_timeout),
|
|
<a href=#1855 id=1855 data-nosnippet>1855</a> <span class="self">self</span>.config.batch_opts.experimental_max_filter_num_per_query,
|
|
<a href=#1856 id=1856 data-nosnippet>1856</a> prefer_short_incremental_cadence,
|
|
<a href=#1857 id=1857 data-nosnippet>1857</a> )
|
|
<a href=#1858 id=1858 data-nosnippet>1858</a> };
|
|
<a href=#1859 id=1859 data-nosnippet>1859</a>
|
|
<a href=#1860 id=1860 data-nosnippet>1860</a> tokio::time::sleep_until(sleep_until).<span class="kw">await</span>;
|
|
<a href=#1861 id=1861 data-nosnippet>1861</a> }
|
|
<a href=#1862 id=1862 data-nosnippet>1862</a> <span class="prelude-val">Ok</span>(<span class="prelude-val">None</span>) => {
|
|
<a href=#1863 id=1863 data-nosnippet>1863</a> <span class="macro">debug!</span>(
|
|
<a href=#1864 id=1864 data-nosnippet>1864</a> <span class="string">"Flow id = {:?} found no new data, sleep for {:?} then continue"</span>,
|
|
<a href=#1865 id=1865 data-nosnippet>1865</a> <span class="self">self</span>.config.flow_id, min_refresh
|
|
<a href=#1866 id=1866 data-nosnippet>1866</a> );
|
|
<a href=#1867 id=1867 data-nosnippet>1867</a> tokio::time::sleep(min_refresh).<span class="kw">await</span>;
|
|
<a href=#1868 id=1868 data-nosnippet>1868</a> <span class="kw">continue</span>;
|
|
<a href=#1869 id=1869 data-nosnippet>1869</a> }
|
|
<a href=#1870 id=1870 data-nosnippet>1870</a> <span class="prelude-val">Err</span>(err) => {
|
|
<a href=#1871 id=1871 data-nosnippet>1871</a> METRIC_FLOW_BATCHING_ENGINE_ERROR_CNT
|
|
<a href=#1872 id=1872 data-nosnippet>1872</a> .with_label_values(<span class="kw-2">&</span>[<span class="kw-2">&</span>flow_id_str])
|
|
<a href=#1873 id=1873 data-nosnippet>1873</a> .inc();
|
|
<a href=#1874 id=1874 data-nosnippet>1874</a> <span class="kw">match </span>outcome.new_query {
|
|
<a href=#1875 id=1875 data-nosnippet>1875</a> <span class="prelude-val">Some</span>(query) => {
|
|
<a href=#1876 id=1876 data-nosnippet>1876</a> <span class="macro">common_telemetry::error!</span>(err; <span class="string">"Failed to execute query for flow={} with query: {}"</span>, <span class="self">self</span>.config.flow_id, query.plan);
|
|
<a href=#1877 id=1877 data-nosnippet>1877</a> max_window_cnt = <span class="prelude-val">Some</span>(<span class="number">1</span>);
|
|
<a href=#1878 id=1878 data-nosnippet>1878</a> }
|
|
<a href=#1879 id=1879 data-nosnippet>1879</a> <span class="prelude-val">None </span>=> {
|
|
<a href=#1880 id=1880 data-nosnippet>1880</a> <span class="macro">common_telemetry::error!</span>(err; <span class="string">"Failed to generate query for flow={}"</span>, <span class="self">self</span>.config.flow_id)
|
|
<a href=#1881 id=1881 data-nosnippet>1881</a> }
|
|
<a href=#1882 id=1882 data-nosnippet>1882</a> }
|
|
<a href=#1883 id=1883 data-nosnippet>1883</a> tokio::time::sleep(min_refresh).<span class="kw">await</span>;
|
|
<a href=#1884 id=1884 data-nosnippet>1884</a> }
|
|
<a href=#1885 id=1885 data-nosnippet>1885</a> }
|
|
<a href=#1886 id=1886 data-nosnippet>1886</a> }
|
|
<a href=#1887 id=1887 data-nosnippet>1887</a> }
|
|
<a href=#1888 id=1888 data-nosnippet>1888</a>
|
|
<a href=#1889 id=1889 data-nosnippet>1889</a> <span class="doccomment">/// Check whether the shutdown signal has been received.
|
|
<a href=#1890 id=1890 data-nosnippet>1890</a> </span><span class="kw">fn </span>is_shutdown_signaled(<span class="kw-2">&</span><span class="self">self</span>) -> bool {
|
|
<a href=#1891 id=1891 data-nosnippet>1891</a> <span class="kw">let </span><span class="kw-2">mut </span>state = <span class="self">self</span>.state.write().unwrap();
|
|
<a href=#1892 id=1892 data-nosnippet>1892</a> <span class="kw">match </span>state.shutdown_rx.try_recv() {
|
|
<a href=#1893 id=1893 data-nosnippet>1893</a> <span class="prelude-val">Ok</span>(()) | <span class="prelude-val">Err</span>(TryRecvError::Closed) => <span class="bool-val">true</span>,
|
|
<a href=#1894 id=1894 data-nosnippet>1894</a> <span class="prelude-val">Err</span>(TryRecvError::Empty) => <span class="bool-val">false</span>,
|
|
<a href=#1895 id=1895 data-nosnippet>1895</a> }
|
|
<a href=#1896 id=1896 data-nosnippet>1896</a> }
|
|
<a href=#1897 id=1897 data-nosnippet>1897</a>
|
|
<a href=#1898 id=1898 data-nosnippet>1898</a> <span class="doccomment">/// Execute one scheduled attempt, temporarily setting
|
|
<a href=#1899 id=1899 data-nosnippet>1899</a> /// `flow.scheduled_time_millis` on the task's QueryContext so
|
|
<a href=#1900 id=1900 data-nosnippet>1900</a> /// SQL/TQL `now()` resolves to the logical scheduled time.
|
|
<a href=#1901 id=1901 data-nosnippet>1901</a> ///
|
|
<a href=#1902 id=1902 data-nosnippet>1902</a> /// The extension is removed after the attempt so a later manual
|
|
<a href=#1903 id=1903 data-nosnippet>1903</a> /// `flush_flow` does not reuse a stale scheduled time.
|
|
<a href=#1904 id=1904 data-nosnippet>1904</a> </span><span class="kw">async fn </span>execute_once_serialized_at_scheduled_time(
|
|
<a href=#1905 id=1905 data-nosnippet>1905</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1906 id=1906 data-nosnippet>1906</a> engine: <span class="kw-2">&</span>QueryEngineRef,
|
|
<a href=#1907 id=1907 data-nosnippet>1907</a> frontend_client: <span class="kw-2">&</span>Arc<FrontendClient>,
|
|
<a href=#1908 id=1908 data-nosnippet>1908</a> scheduled_time_secs: i64,
|
|
<a href=#1909 id=1909 data-nosnippet>1909</a> ) -> ExecuteOnceOutcome {
|
|
<a href=#1910 id=1910 data-nosnippet>1910</a> <span class="kw">let </span><span class="kw-2">mut </span>guard = BatchingExecutionGuard::new(<span class="self">self</span>.execution_lock.clone().lock_owned().<span class="kw">await</span>);
|
|
<a href=#1911 id=1911 data-nosnippet>1911</a>
|
|
<a href=#1912 id=1912 data-nosnippet>1912</a> <span class="comment">// Convert to milliseconds before touching the task state so an
|
|
<a href=#1913 id=1913 data-nosnippet>1913</a> // unrepresentable scheduled time fails as an explicit error without
|
|
<a href=#1914 id=1914 data-nosnippet>1914</a> // ever installing a saturated (off-phase) extension value.
|
|
<a href=#1915 id=1915 data-nosnippet>1915</a> </span><span class="kw">let </span>scheduled_time_millis = <span class="kw">match </span>scheduled_time_millis(scheduled_time_secs) {
|
|
<a href=#1916 id=1916 data-nosnippet>1916</a> <span class="prelude-val">Ok</span>(millis) => millis,
|
|
<a href=#1917 id=1917 data-nosnippet>1917</a> <span class="prelude-val">Err</span>(e) => {
|
|
<a href=#1918 id=1918 data-nosnippet>1918</a> <span class="kw">return </span>ExecuteOnceOutcome {
|
|
<a href=#1919 id=1919 data-nosnippet>1919</a> new_query: <span class="prelude-val">None</span>,
|
|
<a href=#1920 id=1920 data-nosnippet>1920</a> result: <span class="prelude-val">Err</span>(e),
|
|
<a href=#1921 id=1921 data-nosnippet>1921</a> };
|
|
<a href=#1922 id=1922 data-nosnippet>1922</a> }
|
|
<a href=#1923 id=1923 data-nosnippet>1923</a> };
|
|
<a href=#1924 id=1924 data-nosnippet>1924</a>
|
|
<a href=#1925 id=1925 data-nosnippet>1925</a> <span class="comment">// Clone the current QueryContext and add the scheduled time
|
|
<a href=#1926 id=1926 data-nosnippet>1926</a> // extension, then swap it into the task state for this attempt.
|
|
<a href=#1927 id=1927 data-nosnippet>1927</a> </span><span class="kw">let </span>old_ctx = {
|
|
<a href=#1928 id=1928 data-nosnippet>1928</a> <span class="kw">let </span><span class="kw-2">mut </span>state = <span class="self">self</span>.state.write().unwrap();
|
|
<a href=#1929 id=1929 data-nosnippet>1929</a> <span class="kw">let </span>old = state.query_ctx.clone();
|
|
<a href=#1930 id=1930 data-nosnippet>1930</a> <span class="kw">let </span><span class="kw-2">mut </span>new_ctx = (<span class="kw-2">*</span>old).clone();
|
|
<a href=#1931 id=1931 data-nosnippet>1931</a> new_ctx.set_extension(
|
|
<a href=#1932 id=1932 data-nosnippet>1932</a> query::options::FLOW_SCHEDULED_TIME_MILLIS,
|
|
<a href=#1933 id=1933 data-nosnippet>1933</a> scheduled_time_millis.to_string(),
|
|
<a href=#1934 id=1934 data-nosnippet>1934</a> );
|
|
<a href=#1935 id=1935 data-nosnippet>1935</a> state.query_ctx = Arc::new(new_ctx);
|
|
<a href=#1936 id=1936 data-nosnippet>1936</a> old
|
|
<a href=#1937 id=1937 data-nosnippet>1937</a> };
|
|
<a href=#1938 id=1938 data-nosnippet>1938</a> guard.restore_query_context(<span class="self">self</span>.state.clone(), old_ctx);
|
|
<a href=#1939 id=1939 data-nosnippet>1939</a>
|
|
<a href=#1940 id=1940 data-nosnippet>1940</a> <span class="comment">// A collaborator may retain the guard in an owned child task. Its Drop
|
|
<a href=#1941 id=1941 data-nosnippet>1941</a> // restores the scheduled context before releasing serialization.
|
|
<a href=#1942 id=1942 data-nosnippet>1942</a> </span><span class="self">self</span>.execute_once_with_guard(guard, engine, frontend_client, <span class="prelude-val">None</span>)
|
|
<a href=#1943 id=1943 data-nosnippet>1943</a> .<span class="kw">await
|
|
<a href=#1944 id=1944 data-nosnippet>1944</a> </span>}
|
|
<a href=#1945 id=1945 data-nosnippet>1945</a>
|
|
<a href=#1946 id=1946 data-nosnippet>1946</a> <span class="doccomment">/// Generate the create table SQL
|
|
<a href=#1947 id=1947 data-nosnippet>1947</a> ///
|
|
<a href=#1948 id=1948 data-nosnippet>1948</a> /// the auto created table will automatically added a `update_at` Milliseconds DEFAULT now() column in the end
|
|
<a href=#1949 id=1949 data-nosnippet>1949</a> /// (for compatibility with flow streaming mode)
|
|
<a href=#1950 id=1950 data-nosnippet>1950</a> ///
|
|
<a href=#1951 id=1951 data-nosnippet>1951</a> /// and it will use first timestamp column as time index, all other columns will be added as normal columns and nullable
|
|
<a href=#1952 id=1952 data-nosnippet>1952</a> </span><span class="kw">async fn </span>gen_create_table_expr(
|
|
<a href=#1953 id=1953 data-nosnippet>1953</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1954 id=1954 data-nosnippet>1954</a> engine: QueryEngineRef,
|
|
<a href=#1955 id=1955 data-nosnippet>1955</a> ) -> <span class="prelude-ty">Result</span><CreateTableExpr, Error> {
|
|
<a href=#1956 id=1956 data-nosnippet>1956</a> <span class="kw">let </span>query_ctx = <span class="self">self</span>.state.read().unwrap().query_ctx.clone();
|
|
<a href=#1957 id=1957 data-nosnippet>1957</a> <span class="kw">let </span>plan =
|
|
<a href=#1958 id=1958 data-nosnippet>1958</a> sql_to_df_plan(query_ctx.clone(), engine.clone(), <span class="kw-2">&</span><span class="self">self</span>.config.query, <span class="bool-val">true</span>).<span class="kw">await</span><span class="question-mark">?</span>;
|
|
<a href=#1959 id=1959 data-nosnippet>1959</a> create_table_with_expr(<span class="kw-2">&</span>plan, <span class="kw-2">&</span><span class="self">self</span>.config.sink_table_name, <span class="kw-2">&</span><span class="self">self</span>.config.query_type)
|
|
<a href=#1960 id=1960 data-nosnippet>1960</a> }
|
|
<a href=#1961 id=1961 data-nosnippet>1961</a>
|
|
<a href=#1962 id=1962 data-nosnippet>1962</a> <span class="doccomment">/// Incremental delta scans are unfiltered by dirty windows; the sequence
|
|
<a href=#1963 id=1963 data-nosnippet>1963</a> /// range, not a time predicate, defines source correctness.
|
|
<a href=#1964 id=1964 data-nosnippet>1964</a> </span><span class="kw">fn </span>should_use_unfiltered_incremental_delta(<span class="kw-2">&</span><span class="self">self</span>) -> bool {
|
|
<a href=#1965 id=1965 data-nosnippet>1965</a> <span class="kw">let </span>state = <span class="self">self</span>.state.read().unwrap();
|
|
<a href=#1966 id=1966 data-nosnippet>1966</a> state.checkpoint_mode() == CheckpointMode::Incremental
|
|
<a href=#1967 id=1967 data-nosnippet>1967</a> && !state.is_incremental_disabled()
|
|
<a href=#1968 id=1968 data-nosnippet>1968</a> && <span class="macro">matches!</span>(<span class="self">self</span>.config.query_type, QueryType::Sql)
|
|
<a href=#1969 id=1969 data-nosnippet>1969</a> }
|
|
<a href=#1970 id=1970 data-nosnippet>1970</a>
|
|
<a href=#1971 id=1971 data-nosnippet>1971</a> <span class="doccomment">/// Generate the next ordinary plan and classify its coverage.
|
|
<a href=#1972 id=1972 data-nosnippet>1972</a> </span><span class="kw">async fn </span>gen_query_with_time_window(
|
|
<a href=#1973 id=1973 data-nosnippet>1973</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1974 id=1974 data-nosnippet>1974</a> engine: QueryEngineRef,
|
|
<a href=#1975 id=1975 data-nosnippet>1975</a> sink_table_schema: <span class="kw-2">&</span>Arc<Schema>,
|
|
<a href=#1976 id=1976 data-nosnippet>1976</a> primary_key_indices: <span class="kw-2">&</span>[usize],
|
|
<a href=#1977 id=1977 data-nosnippet>1977</a> allow_partial: bool,
|
|
<a href=#1978 id=1978 data-nosnippet>1978</a> max_window_cnt: <span class="prelude-ty">Option</span><usize>,
|
|
<a href=#1979 id=1979 data-nosnippet>1979</a> ) -> <span class="prelude-ty">Result</span><<span class="prelude-ty">Option</span><PlanInfo>, Error> {
|
|
<a href=#1980 id=1980 data-nosnippet>1980</a> <span class="self">self</span>.gen_query_with_time_window_with_values(
|
|
<a href=#1981 id=1981 data-nosnippet>1981</a> engine,
|
|
<a href=#1982 id=1982 data-nosnippet>1982</a> sink_table_schema,
|
|
<a href=#1983 id=1983 data-nosnippet>1983</a> primary_key_indices,
|
|
<a href=#1984 id=1984 data-nosnippet>1984</a> allow_partial,
|
|
<a href=#1985 id=1985 data-nosnippet>1985</a> max_window_cnt,
|
|
<a href=#1986 id=1986 data-nosnippet>1986</a> <span class="kw-2">&</span>BTreeMap::new(),
|
|
<a href=#1987 id=1987 data-nosnippet>1987</a> <span class="bool-val">false</span>,
|
|
<a href=#1988 id=1988 data-nosnippet>1988</a> )
|
|
<a href=#1989 id=1989 data-nosnippet>1989</a> .<span class="kw">await
|
|
<a href=#1990 id=1990 data-nosnippet>1990</a> </span>}
|
|
<a href=#1991 id=1991 data-nosnippet>1991</a>
|
|
<a href=#1992 id=1992 data-nosnippet>1992</a> <span class="doccomment">/// Generate the next plan with extension-owned sink values and an optional forced full scan.
|
|
<a href=#1993 id=1993 data-nosnippet>1993</a> </span><span class="attr">#[allow(clippy::too_many_arguments)]
|
|
<a href=#1994 id=1994 data-nosnippet>1994</a> </span><span class="kw">async fn </span>gen_query_with_time_window_with_values(
|
|
<a href=#1995 id=1995 data-nosnippet>1995</a> <span class="kw-2">&</span><span class="self">self</span>,
|
|
<a href=#1996 id=1996 data-nosnippet>1996</a> engine: QueryEngineRef,
|
|
<a href=#1997 id=1997 data-nosnippet>1997</a> sink_table_schema: <span class="kw-2">&</span>Arc<Schema>,
|
|
<a href=#1998 id=1998 data-nosnippet>1998</a> primary_key_indices: <span class="kw-2">&</span>[usize],
|
|
<a href=#1999 id=1999 data-nosnippet>1999</a> allow_partial: bool,
|
|
<a href=#2000 id=2000 data-nosnippet>2000</a> max_window_cnt: <span class="prelude-ty">Option</span><usize>,
|
|
<a href=#2001 id=2001 data-nosnippet>2001</a> values: <span class="kw-2">&</span>BTreeMap<String, ScalarValue>,
|
|
<a href=#2002 id=2002 data-nosnippet>2002</a> force_full_snapshot: bool,
|
|
<a href=#2003 id=2003 data-nosnippet>2003</a> ) -> <span class="prelude-ty">Result</span><<span class="prelude-ty">Option</span><PlanInfo>, Error> {
|
|
<a href=#2004 id=2004 data-nosnippet>2004</a> <span class="kw">let </span>query_ctx = <span class="self">self</span>.state.read().unwrap().query_ctx.clone();
|
|
<a href=#2005 id=2005 data-nosnippet>2005</a> <span class="kw">let </span>start = SystemTime::now();
|
|
<a href=#2006 id=2006 data-nosnippet>2006</a> <span class="kw">let </span>since_the_epoch = start
|
|
<a href=#2007 id=2007 data-nosnippet>2007</a> .duration_since(UNIX_EPOCH)
|
|
<a href=#2008 id=2008 data-nosnippet>2008</a> .expect(<span class="string">"Time went backwards"</span>);
|
|
<a href=#2009 id=2009 data-nosnippet>2009</a> <span class="kw">let </span>low_bound = <span class="self">self
|
|
<a href=#2010 id=2010 data-nosnippet>2010</a> </span>.config
|
|
<a href=#2011 id=2011 data-nosnippet>2011</a> .expire_after
|
|
<a href=#2012 id=2012 data-nosnippet>2012</a> .map(|e| since_the_epoch.as_secs() - e <span class="kw">as </span>u64)
|
|
<a href=#2013 id=2013 data-nosnippet>2013</a> .unwrap_or(u64::MIN);
|
|
<a href=#2014 id=2014 data-nosnippet>2014</a>
|
|
<a href=#2015 id=2015 data-nosnippet>2015</a> <span class="kw">let </span>low_bound = Timestamp::new_second(low_bound <span class="kw">as </span>i64);
|
|
<a href=#2016 id=2016 data-nosnippet>2016</a>
|
|
<a href=#2017 id=2017 data-nosnippet>2017</a> <span class="kw">let </span>expire_time_window_bound = <span class="self">self
|
|
<a href=#2018 id=2018 data-nosnippet>2018</a> </span>.config
|
|
<a href=#2019 id=2019 data-nosnippet>2019</a> .time_window_expr
|
|
<a href=#2020 id=2020 data-nosnippet>2020</a> .as_ref()
|
|
<a href=#2021 id=2021 data-nosnippet>2021</a> .map(|expr| expr.eval(low_bound))
|
|
<a href=#2022 id=2022 data-nosnippet>2022</a> .transpose()<span class="question-mark">?</span>;
|
|
<a href=#2023 id=2023 data-nosnippet>2023</a>
|
|
<a href=#2024 id=2024 data-nosnippet>2024</a> <span class="kw">let </span>(expire_lower_bound, expire_upper_bound) = <span class="kw">match </span>(
|
|
<a href=#2025 id=2025 data-nosnippet>2025</a> expire_time_window_bound,
|
|
<a href=#2026 id=2026 data-nosnippet>2026</a> <span class="kw-2">&</span><span class="self">self</span>.config.query_type,
|
|
<a href=#2027 id=2027 data-nosnippet>2027</a> ) {
|
|
<a href=#2028 id=2028 data-nosnippet>2028</a> (<span class="prelude-val">Some</span>((<span class="prelude-val">Some</span>(l), <span class="prelude-val">Some</span>(u))), QueryType::Sql) => (l, u),
|
|
<a href=#2029 id=2029 data-nosnippet>2029</a> (<span class="prelude-val">None</span>, QueryType::Sql) <span class="kw">if </span><span class="self">self</span>.config.flow_eval_interval.is_none() => {
|
|
<a href=#2030 id=2030 data-nosnippet>2030</a> <span class="kw">return </span>UnexpectedSnafu {
|
|
<a href=#2031 id=2031 data-nosnippet>2031</a> reason: <span class="macro">format!</span>(
|
|
<a href=#2032 id=2032 data-nosnippet>2032</a> <span class="string">"Flow id={} reached execution without a time-window expression or EVAL INTERVAL; create-flow validation should have rejected it"</span>,
|
|
<a href=#2033 id=2033 data-nosnippet>2033</a> <span class="self">self</span>.config.flow_id
|
|
<a href=#2034 id=2034 data-nosnippet>2034</a> ),
|
|
<a href=#2035 id=2035 data-nosnippet>2035</a> }
|
|
<a href=#2036 id=2036 data-nosnippet>2036</a> .fail();
|
|
<a href=#2037 id=2037 data-nosnippet>2037</a> }
|
|
<a href=#2038 id=2038 data-nosnippet>2038</a> <span class="kw">_ </span>=> {
|
|
<a href=#2039 id=2039 data-nosnippet>2039</a> <span class="comment">// Explicit full-query flows (TQL and evaluation-interval SQL
|
|
<a href=#2040 id=2040 data-nosnippet>2040</a> // plans whose shape cannot be safely dirty-window pruned) are
|
|
<a href=#2041 id=2041 data-nosnippet>2041</a> // allowed to run as unfiltered full snapshots. This is distinct
|
|
<a href=#2042 id=2042 data-nosnippet>2042</a> // from using unfiltered full as a fallback after scoped repair or
|
|
<a href=#2043 id=2043 data-nosnippet>2043</a> // incremental rewrite failed.
|
|
<a href=#2044 id=2044 data-nosnippet>2044</a> </span><span class="kw">let </span>(<span class="kw">_</span>, dirty_windows_to_restore) = <span class="self">self</span>.drain_dirty_windows_signal();
|
|
<a href=#2045 id=2045 data-nosnippet>2045</a>
|
|
<a href=#2046 id=2046 data-nosnippet>2046</a> <span class="kw">let </span>plan_info = <span class="self">self
|
|
<a href=#2047 id=2047 data-nosnippet>2047</a> </span>.gen_unfiltered_plan_info(
|
|
<a href=#2048 id=2048 data-nosnippet>2048</a> engine,
|
|
<a href=#2049 id=2049 data-nosnippet>2049</a> query_ctx,
|
|
<a href=#2050 id=2050 data-nosnippet>2050</a> sink_table_schema.clone(),
|
|
<a href=#2051 id=2051 data-nosnippet>2051</a> primary_key_indices,
|
|
<a href=#2052 id=2052 data-nosnippet>2052</a> allow_partial,
|
|
<a href=#2053 id=2053 data-nosnippet>2053</a> values,
|
|
<a href=#2054 id=2054 data-nosnippet>2054</a> dirty_windows_to_restore,
|
|
<a href=#2055 id=2055 data-nosnippet>2055</a> <span class="prelude-val">None</span>,
|
|
<a href=#2056 id=2056 data-nosnippet>2056</a> QueryCoverage::UnfilteredFull,
|
|
<a href=#2057 id=2057 data-nosnippet>2057</a> )
|
|
<a href=#2058 id=2058 data-nosnippet>2058</a> .<span class="kw">await</span><span class="question-mark">?</span>;
|
|
<a href=#2059 id=2059 data-nosnippet>2059</a>
|
|
<a href=#2060 id=2060 data-nosnippet>2060</a> <span class="kw">return </span><span class="prelude-val">Ok</span>(<span class="prelude-val">Some</span>(plan_info));
|
|
<a href=#2061 id=2061 data-nosnippet>2061</a> }
|
|
<a href=#2062 id=2062 data-nosnippet>2062</a> };
|
|
<a href=#2063 id=2063 data-nosnippet>2063</a>
|
|
<a href=#2064 id=2064 data-nosnippet>2064</a> <span class="macro">debug!</span>(
|
|
<a href=#2065 id=2065 data-nosnippet>2065</a> <span class="string">"Flow id = {:?}, found time window: precise_lower_bound={:?}, precise_upper_bound={:?} with dirty time windows: {:?}"</span>,
|
|
<a href=#2066 id=2066 data-nosnippet>2066</a> <span class="self">self</span>.config.flow_id,
|
|
<a href=#2067 id=2067 data-nosnippet>2067</a> expire_lower_bound,
|
|
<a href=#2068 id=2068 data-nosnippet>2068</a> expire_upper_bound,
|
|
<a href=#2069 id=2069 data-nosnippet>2069</a> <span class="self">self</span>.state.read().unwrap().dirty_time_windows
|
|
<a href=#2070 id=2070 data-nosnippet>2070</a> );
|
|
<a href=#2071 id=2071 data-nosnippet>2071</a> <span class="kw">let </span>window_size = expire_upper_bound
|
|
<a href=#2072 id=2072 data-nosnippet>2072</a> .sub(<span class="kw-2">&</span>expire_lower_bound)
|
|
<a href=#2073 id=2073 data-nosnippet>2073</a> .with_context(|| UnexpectedSnafu {
|
|
<a href=#2074 id=2074 data-nosnippet>2074</a> reason: <span class="macro">format!</span>(
|
|
<a href=#2075 id=2075 data-nosnippet>2075</a> <span class="string">"Can't get window size from {expire_upper_bound:?} - {expire_lower_bound:?}"
|
|
<a href=#2076 id=2076 data-nosnippet>2076</a> </span>),
|
|
<a href=#2077 id=2077 data-nosnippet>2077</a> })<span class="question-mark">?</span>;
|
|
<a href=#2078 id=2078 data-nosnippet>2078</a> <span class="kw">let </span>col_name = <span class="self">self
|
|
<a href=#2079 id=2079 data-nosnippet>2079</a> </span>.config
|
|
<a href=#2080 id=2080 data-nosnippet>2080</a> .time_window_expr
|
|
<a href=#2081 id=2081 data-nosnippet>2081</a> .as_ref()
|
|
<a href=#2082 id=2082 data-nosnippet>2082</a> .map(|expr| expr.column_name.clone())
|
|
<a href=#2083 id=2083 data-nosnippet>2083</a> .with_context(|| UnexpectedSnafu {
|
|
<a href=#2084 id=2084 data-nosnippet>2084</a> reason: <span class="macro">format!</span>(
|
|
<a href=#2085 id=2085 data-nosnippet>2085</a> <span class="string">"Flow id={:?}, Failed to get column name from time window expr"</span>,
|
|
<a href=#2086 id=2086 data-nosnippet>2086</a> <span class="self">self</span>.config.flow_id
|
|
<a href=#2087 id=2087 data-nosnippet>2087</a> ),
|
|
<a href=#2088 id=2088 data-nosnippet>2088</a> })<span class="question-mark">?</span>;
|
|
<a href=#2089 id=2089 data-nosnippet>2089</a>
|
|
<a href=#2090 id=2090 data-nosnippet>2090</a> <span class="kw">if </span>force_full_snapshot {
|
|
<a href=#2091 id=2091 data-nosnippet>2091</a> <span class="comment">// Durable execution cannot trust an uncertain prior checkpoint. Re-run an
|
|
<a href=#2092 id=2092 data-nosnippet>2092</a> // unfiltered full snapshot, but retain the normal expiry predicate and the consumed
|
|
<a href=#2093 id=2093 data-nosnippet>2093</a> // dirty signal so every planning/execution failure remains reversible.
|
|
<a href=#2094 id=2094 data-nosnippet>2094</a> </span><span class="kw">let </span>retention_filter = <span class="self">self</span>.config.expire_after.map(|<span class="kw">_</span>| {
|
|
<a href=#2095 id=2095 data-nosnippet>2095</a> (
|
|
<a href=#2096 id=2096 data-nosnippet>2096</a> col_name.as_str(),
|
|
<a href=#2097 id=2097 data-nosnippet>2097</a> expire_lower_bound,
|
|
<a href=#2098 id=2098 data-nosnippet>2098</a> <span class="string">"forced full snapshot"</span>,
|
|
<a href=#2099 id=2099 data-nosnippet>2099</a> )
|
|
<a href=#2100 id=2100 data-nosnippet>2100</a> });
|
|
<a href=#2101 id=2101 data-nosnippet>2101</a> <span class="kw">let </span>(<span class="kw">_</span>, dirty_windows_to_restore) = <span class="self">self</span>.drain_dirty_windows_signal();
|
|
<a href=#2102 id=2102 data-nosnippet>2102</a> <span class="kw">return </span><span class="self">self
|
|
<a href=#2103 id=2103 data-nosnippet>2103</a> </span>.gen_unfiltered_plan_info(
|
|
<a href=#2104 id=2104 data-nosnippet>2104</a> engine,
|
|
<a href=#2105 id=2105 data-nosnippet>2105</a> query_ctx,
|
|
<a href=#2106 id=2106 data-nosnippet>2106</a> sink_table_schema.clone(),
|
|
<a href=#2107 id=2107 data-nosnippet>2107</a> primary_key_indices,
|
|
<a href=#2108 id=2108 data-nosnippet>2108</a> allow_partial,
|
|
<a href=#2109 id=2109 data-nosnippet>2109</a> values,
|
|
<a href=#2110 id=2110 data-nosnippet>2110</a> dirty_windows_to_restore,
|
|
<a href=#2111 id=2111 data-nosnippet>2111</a> retention_filter,
|
|
<a href=#2112 id=2112 data-nosnippet>2112</a> QueryCoverage::UnfilteredFull,
|
|
<a href=#2113 id=2113 data-nosnippet>2113</a> )
|
|
<a href=#2114 id=2114 data-nosnippet>2114</a> .<span class="kw">await
|
|
<a href=#2115 id=2115 data-nosnippet>2115</a> </span>.map(<span class="prelude-val">Some</span>);
|
|
<a href=#2116 id=2116 data-nosnippet>2116</a> }
|
|
<a href=#2117 id=2117 data-nosnippet>2117</a>
|
|
<a href=#2118 id=2118 data-nosnippet>2118</a> <span class="kw">if </span><span class="self">self</span>.should_use_unfiltered_incremental_delta() {
|
|
<a href=#2119 id=2119 data-nosnippet>2119</a> <span class="comment">// In incremental mode, source correctness is defined by the
|
|
<a href=#2120 id=2120 data-nosnippet>2120</a> // per-region sequence range `(checkpoint, scan-open snapshot]`, not
|
|
<a href=#2121 id=2121 data-nosnippet>2121</a> // by dirty-window predicates. Dirty windows are only a scheduling
|
|
<a href=#2122 id=2122 data-nosnippet>2122</a> // signal here. Applying a stale dirty-window filter to the source can
|
|
<a href=#2123 id=2123 data-nosnippet>2123</a> // exclude rows that are inside the returned watermark and make a
|
|
<a href=#2124 id=2124 data-nosnippet>2124</a> // checkpoint advance skip them forever. The sink side is also left
|
|
<a href=#2125 id=2125 data-nosnippet>2125</a> // unfiltered by dirty windows; the incremental rewrite joins the
|
|
<a href=#2126 id=2126 data-nosnippet>2126</a> // delta groups with the full sink state for correctness. Future
|
|
<a href=#2127 id=2127 data-nosnippet>2127</a> // dynamic filters can prune sink reads as a pure optimization.
|
|
<a href=#2128 id=2128 data-nosnippet>2128</a> </span><span class="kw">let </span>retention_filter = <span class="self">self
|
|
<a href=#2129 id=2129 data-nosnippet>2129</a> </span>.config
|
|
<a href=#2130 id=2130 data-nosnippet>2130</a> .expire_after
|
|
<a href=#2131 id=2131 data-nosnippet>2131</a> .map(|<span class="kw">_</span>| (col_name.as_str(), expire_lower_bound, <span class="string">"incremental"</span>));
|
|
<a href=#2132 id=2132 data-nosnippet>2132</a> <span class="kw">return </span><span class="self">self
|
|
<a href=#2133 id=2133 data-nosnippet>2133</a> </span>.gen_unfiltered_plan_info_if_dirty(
|
|
<a href=#2134 id=2134 data-nosnippet>2134</a> engine,
|
|
<a href=#2135 id=2135 data-nosnippet>2135</a> query_ctx,
|
|
<a href=#2136 id=2136 data-nosnippet>2136</a> sink_table_schema.clone(),
|
|
<a href=#2137 id=2137 data-nosnippet>2137</a> primary_key_indices,
|
|
<a href=#2138 id=2138 data-nosnippet>2138</a> allow_partial,
|
|
<a href=#2139 id=2139 data-nosnippet>2139</a> values,
|
|
<a href=#2140 id=2140 data-nosnippet>2140</a> retention_filter,
|
|
<a href=#2141 id=2141 data-nosnippet>2141</a> QueryCoverage::IncrementalDelta,
|
|
<a href=#2142 id=2142 data-nosnippet>2142</a> )
|
|
<a href=#2143 id=2143 data-nosnippet>2143</a> .<span class="kw">await</span>;
|
|
<a href=#2144 id=2144 data-nosnippet>2144</a> }
|
|
<a href=#2145 id=2145 data-nosnippet>2145</a>
|
|
<a href=#2146 id=2146 data-nosnippet>2146</a> <span class="kw">let </span>(expr, coverage) = {
|
|
<a href=#2147 id=2147 data-nosnippet>2147</a> <span class="kw">let </span><span class="kw-2">mut </span>state = <span class="self">self</span>.state.write().unwrap();
|
|
<a href=#2148 id=2148 data-nosnippet>2148</a> <span class="kw">let </span>window_cnt = max_window_cnt
|
|
<a href=#2149 id=2149 data-nosnippet>2149</a> .unwrap_or(<span class="self">self</span>.config.batch_opts.experimental_max_filter_num_per_query);
|
|
<a href=#2150 id=2150 data-nosnippet>2150</a> <span class="kw">let </span>expr = state.gen_scoped_filter_exprs(
|
|
<a href=#2151 id=2151 data-nosnippet>2151</a> <span class="kw-2">&</span>col_name,
|
|
<a href=#2152 id=2152 data-nosnippet>2152</a> <span class="prelude-val">Some</span>(expire_lower_bound),
|
|
<a href=#2153 id=2153 data-nosnippet>2153</a> window_size,
|
|
<a href=#2154 id=2154 data-nosnippet>2154</a> window_cnt,
|
|
<a href=#2155 id=2155 data-nosnippet>2155</a> <span class="self">self</span>.config.flow_id,
|
|
<a href=#2156 id=2156 data-nosnippet>2156</a> <span class="prelude-val">Some</span>(<span class="self">self</span>),
|
|
<a href=#2157 id=2157 data-nosnippet>2157</a> )<span class="question-mark">?</span>;
|
|
<a href=#2158 id=2158 data-nosnippet>2158</a> <span class="kw">let </span>repair_high = state
|
|
<a href=#2159 id=2159 data-nosnippet>2159</a> .pending_fenced_repair()
|
|
<a href=#2160 id=2160 data-nosnippet>2160</a> .map(|repair| repair.high().clone());
|
|
<a href=#2161 id=2161 data-nosnippet>2161</a> <span class="kw">let </span>coverage = <span class="kw">if let </span><span class="prelude-val">Some</span>(high) = repair_high {
|
|
<a href=#2162 id=2162 data-nosnippet>2162</a> QueryCoverage::FencedRepairChunk { high }
|
|
<a href=#2163 id=2163 data-nosnippet>2163</a> } <span class="kw">else </span>{
|
|
<a href=#2164 id=2164 data-nosnippet>2164</a> QueryCoverage::ScopedBaseRepair
|
|
<a href=#2165 id=2165 data-nosnippet>2165</a> };
|
|
<a href=#2166 id=2166 data-nosnippet>2166</a> (expr, coverage)
|
|
<a href=#2167 id=2167 data-nosnippet>2167</a> };
|
|
<a href=#2168 id=2168 data-nosnippet>2168</a>
|
|
<a href=#2169 id=2169 data-nosnippet>2169</a> <span class="kw">let </span><span class="prelude-val">Some</span>(expr) = expr <span class="kw">else </span>{
|
|
<a href=#2170 id=2170 data-nosnippet>2170</a> <span class="comment">// no new data, hence no need to update
|
|
<a href=#2171 id=2171 data-nosnippet>2171</a> </span><span class="macro">debug!</span>(<span class="string">"Flow id={:?}, no new data, not update"</span>, <span class="self">self</span>.config.flow_id);
|
|
<a href=#2172 id=2172 data-nosnippet>2172</a> <span class="kw">return </span><span class="prelude-val">Ok</span>(<span class="prelude-val">None</span>);
|
|
<a href=#2173 id=2173 data-nosnippet>2173</a> };
|
|
<a href=#2174 id=2174 data-nosnippet>2174</a>
|
|
<a href=#2175 id=2175 data-nosnippet>2175</a> <span class="kw">let </span>filter_sql = expr_to_sql(<span class="kw-2">&</span>expr.expr)
|
|
<a href=#2176 id=2176 data-nosnippet>2176</a> .map(|sql| sql.to_string())
|
|
<a href=#2177 id=2177 data-nosnippet>2177</a> .unwrap_or_else(|err| <span class="macro">format!</span>(<span class="string">"<failed to format filter expr: {err}>"</span>));
|
|
<a href=#2178 id=2178 data-nosnippet>2178</a>
|
|
<a href=#2179 id=2179 data-nosnippet>2179</a> <span class="macro">debug!</span>(
|
|
<a href=#2180 id=2180 data-nosnippet>2180</a> <span class="string">"Flow id={:?}, Generated filter expr: {:?}"</span>,
|
|
<a href=#2181 id=2181 data-nosnippet>2181</a> <span class="self">self</span>.config.flow_id, filter_sql
|
|
<a href=#2182 id=2182 data-nosnippet>2182</a> );
|
|
<a href=#2183 id=2183 data-nosnippet>2183</a>
|
|
<a href=#2184 id=2184 data-nosnippet>2184</a> <span class="kw">let </span><span class="kw-2">mut </span>add_filter = AddFilterRewriter::new(expr.expr.clone());
|
|
<a href=#2185 id=2185 data-nosnippet>2185</a> <span class="kw">let </span><span class="kw-2">mut </span>add_auto_column = ColumnMatcherRewriter::new_with_values(
|
|
<a href=#2186 id=2186 data-nosnippet>2186</a> sink_table_schema.clone(),
|
|
<a href=#2187 id=2187 data-nosnippet>2187</a> primary_key_indices.to_vec(),
|
|
<a href=#2188 id=2188 data-nosnippet>2188</a> allow_partial,
|
|
<a href=#2189 id=2189 data-nosnippet>2189</a> values.clone(),
|
|
<a href=#2190 id=2190 data-nosnippet>2190</a> );
|
|
<a href=#2191 id=2191 data-nosnippet>2191</a>
|
|
<a href=#2192 id=2192 data-nosnippet>2192</a> <span class="kw">let </span>plan = <span class="self">self</span>.restore_scoped_dirty_windows_on_err(
|
|
<a href=#2193 id=2193 data-nosnippet>2193</a> <span class="kw-2">&</span>expr,
|
|
<a href=#2194 id=2194 data-nosnippet>2194</a> sql_to_df_plan(query_ctx.clone(), engine.clone(), <span class="kw-2">&</span><span class="self">self</span>.config.query, <span class="bool-val">false</span>).<span class="kw">await</span>,
|
|
<a href=#2195 id=2195 data-nosnippet>2195</a> )<span class="question-mark">?</span>;
|
|
<a href=#2196 id=2196 data-nosnippet>2196</a> <span class="kw">let </span>rewrite = <span class="self">self</span>.restore_scoped_dirty_windows_on_err(
|
|
<a href=#2197 id=2197 data-nosnippet>2197</a> <span class="kw-2">&</span>expr,
|
|
<a href=#2198 id=2198 data-nosnippet>2198</a> plan.clone()
|
|
<a href=#2199 id=2199 data-nosnippet>2199</a> .rewrite(<span class="kw-2">&mut </span>add_filter)
|
|
<a href=#2200 id=2200 data-nosnippet>2200</a> .and_then(|p| p.data.rewrite(<span class="kw-2">&mut </span>add_auto_column))
|
|
<a href=#2201 id=2201 data-nosnippet>2201</a> .with_context(|<span class="kw">_</span>| DatafusionSnafu {
|
|
<a href=#2202 id=2202 data-nosnippet>2202</a> context: <span class="macro">format!</span>(<span class="string">"Failed to rewrite plan:\n {}\n"</span>, plan),
|
|
<a href=#2203 id=2203 data-nosnippet>2203</a> })
|
|
<a href=#2204 id=2204 data-nosnippet>2204</a> .map(|rewrite| rewrite.data),
|
|
<a href=#2205 id=2205 data-nosnippet>2205</a> )<span class="question-mark">?</span>;
|
|
<a href=#2206 id=2206 data-nosnippet>2206</a> <span class="comment">// only apply optimize after complex rewrite is done
|
|
<a href=#2207 id=2207 data-nosnippet>2207</a> </span><span class="kw">let </span>new_plan = <span class="self">self</span>.restore_scoped_dirty_windows_on_err(
|
|
<a href=#2208 id=2208 data-nosnippet>2208</a> <span class="kw-2">&</span>expr,
|
|
<a href=#2209 id=2209 data-nosnippet>2209</a> apply_df_optimizer(rewrite, <span class="kw-2">&</span>query_ctx).<span class="kw">await</span>,
|
|
<a href=#2210 id=2210 data-nosnippet>2210</a> )<span class="question-mark">?</span>;
|
|
<a href=#2211 id=2211 data-nosnippet>2211</a>
|
|
<a href=#2212 id=2212 data-nosnippet>2212</a> <span class="kw">let </span>info = PlanInfo {
|
|
<a href=#2213 id=2213 data-nosnippet>2213</a> plan: new_plan.clone(),
|
|
<a href=#2214 id=2214 data-nosnippet>2214</a> dirty_restore: DirtyRestore::Scoped(expr),
|
|
<a href=#2215 id=2215 data-nosnippet>2215</a> coverage,
|
|
<a href=#2216 id=2216 data-nosnippet>2216</a> };
|
|
<a href=#2217 id=2217 data-nosnippet>2217</a>
|
|
<a href=#2218 id=2218 data-nosnippet>2218</a> <span class="prelude-val">Ok</span>(<span class="prelude-val">Some</span>(info))
|
|
<a href=#2219 id=2219 data-nosnippet>2219</a> }
|
|
<a href=#2220 id=2220 data-nosnippet>2220</a>}
|
|
<a href=#2221 id=2221 data-nosnippet>2221</a>
|
|
<a href=#2222 id=2222 data-nosnippet>2222</a><span class="attr">#[cfg(test)]
|
|
<a href=#2223 id=2223 data-nosnippet>2223</a></span><span class="kw">mod </span>test;
|
|
</code></pre></div></section></main></body></html> |