[
https://issues.apache.org/jira/browse/CALCITE-7251?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mihai Budiu updated CALCITE-7251:
---------------------------------
Description:
In [CALCITE-6071] I have added support for some Rex nodes to carry source
information position. I propose adding this to more kinds of nodes: SEARCH and
WINDOW are two examples that are frequently used and easy to modify. This is
really a kind of debugging information, and it can help provide better runtime
diagnostics.
SEARCH nodes are tricky, because by construction they combine information from
multiple expressions; the proposal is to have one position per search which is
the SUM of the positions of the component expressions. This leads to some
information loss, but it's better than nothing, and in practice the components
are very likely to come from nearby positions in the source code.
was:
In [CALCITE-6701] I have added support for some Rex nodes to carry source
information position. I propose adding this to more kinds of nodes: SEARCH and
WINDOW are two examples that are frequently used and easy to modify. This is
really a kind of debugging information, and it can help provide better runtime
diagnostics.
SEARCH nodes are tricky, because by construction they combine information from
multiple expressions; the proposal is to have one position per search which is
the SUM of the positions of the component expressions. This leads to some
information loss, but it's better than nothing, and in practice the components
are very likely to come from nearby positions in the source code.
> SEARCH and WINDOW operations lose source position information
> -------------------------------------------------------------
>
> Key: CALCITE-7251
> URL: https://issues.apache.org/jira/browse/CALCITE-7251
> Project: Calcite
> Issue Type: Improvement
> Components: core
> Affects Versions: 1.40.0
> Reporter: Mihai Budiu
> Assignee: Mihai Budiu
> Priority: Minor
>
> In [CALCITE-6071] I have added support for some Rex nodes to carry source
> information position. I propose adding this to more kinds of nodes: SEARCH
> and WINDOW are two examples that are frequently used and easy to modify. This
> is really a kind of debugging information, and it can help provide better
> runtime diagnostics.
> SEARCH nodes are tricky, because by construction they combine information
> from multiple expressions; the proposal is to have one position per search
> which is the SUM of the positions of the component expressions. This leads to
> some information loss, but it's better than nothing, and in practice the
> components are very likely to come from nearby positions in the source code.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
