SQL: rewrite ROUND and TRUNCATE functions with a different optional parameter handling method #40242
Conversation
approach to handling the optional parameter in the constructor. Until now the optional parameter was considered 0 if the value was missing and the constructor was filling in this value. The current solution is to have the optional parameter as null right until the actual calculation is done.
astefan
added
>bug
v7.0.0
:Analytics/SQL
|
Pinging @elastic/es-search |
|
|
||
| } | ||
|
|
||
| return right == null ? TypeResolution.TYPE_RESOLVED : isNumeric(right, sourceText(), ParamOrdinal.SECOND); |
There was a problem hiding this comment.
The right should be of integer type (not rational), so we should catch it here before the execution time.
| public DataType dataType() { | ||
| return left().dataType(); | ||
| public Expression replaceChildren(List<Expression> newChildren) { | ||
| if (right() != null && newChildren.size() != 2) { |
There was a problem hiding this comment.
I'd extract this to the parent method and leave here only the instance creation, as it's duplicated also in Round.
| } | ||
|
|
||
| private ScriptTemplate asScriptFrom(ScriptTemplate leftScript, ScriptTemplate rightScript) { | ||
| if (right == null) { |
There was a problem hiding this comment.
Wouldn't be possible to avoid the "double" methods in painless and just have one with both args where null is passed for the right arg?
There was a problem hiding this comment.
Gave it a try, but failed. If you have any ideas, I'd love to hear them. Otherwise, I'll try later today or tomorrow.
There was a problem hiding this comment.
I took some time here and it works like that: matriv@b35090e
There was a problem hiding this comment.
Indeed. Nice one. I've incorporated it in the PR. Thanks.
| import java.util.function.BiFunction; | ||
|
|
||
| /** | ||
| * Processor for binary mathematical operations that can have the second parameter optional. |
There was a problem hiding this comment.
Maybe rephrase: ... that have a second optional parameter.
| @@ -200,6 +200,23 @@ ROUND(SQRT(CAST(EXP(languages) AS SMALLINT)),2):d| COUNT(*):l | |||
| null |10 | |||
| ; | |||
|
|
|||
| groupByRoundWithTwoParams | |||
There was a problem hiding this comment.
Can't this be tested in QueryTranslatorTests?
|
@elasticmachine run elasticsearch-ci/1 |
|
@elasticmachine run elasticsearch-ci/default-distro |
3 similar comments
|
@elasticmachine run elasticsearch-ci/default-distro |
|
@elasticmachine run elasticsearch-ci/default-distro |
|
@elasticmachine run elasticsearch-ci/default-distro |
…arameter handling method (#40242) * Rewrite Round and Truncate functions to have a slightly different approach to handling the optional parameter in the constructor. Until now the optional parameter was considered 0 if the value was missing and the constructor was filling in this value. The current solution is to have the optional parameter as null right until the actual calculation is done. (cherry picked from commit 3e314f8)
…arameter handling method (#40242) * Rewrite Round and Truncate functions to have a slightly different approach to handling the optional parameter in the constructor. Until now the optional parameter was considered 0 if the value was missing and the constructor was filling in this value. The current solution is to have the optional parameter as null right until the actual calculation is done. (cherry picked from commit 3e314f8)
…arameter handling method (#40242) * Rewrite Round and Truncate functions to have a slightly different approach to handling the optional parameter in the constructor. Until now the optional parameter was considered 0 if the value was missing and the constructor was filling in this value. The current solution is to have the optional parameter as null right until the actual calculation is done. (cherry picked from commit 3e314f8)
…arameter handling method (#40242) * Rewrite Round and Truncate functions to have a slightly different approach to handling the optional parameter in the constructor. Until now the optional parameter was considered 0 if the value was missing and the constructor was filling in this value. The current solution is to have the optional parameter as null right until the actual calculation is done. (cherry picked from commit 3e314f8)
The optional parameter for these two functions was handled in the constructor by passing the value of
0whenever the parameter was missing. This had an apparently minor implication of defining a second child for the function (the second parameter). If the value was kept asnull(missing) the function would have had a single child (the first parameter).This affected the comparison operation from the Analyzer and the same function that was already resolved wouldn't match the new function found because the new one had no second parameter while the resolved one had the parameter
0. Because of this comparison failing, aGROUP BYthe same function would fail.This PR also fixes a bug with ATAN2 and POWER functions which didn't have any Painless methods.
Fixes #40001.