Normalize index identifiers on construction#142
Conversation
Index column identifiers are stripped of outer quotes at construction. Indexes created by older versions still validate via a reconstructed legacy name candidate.
3f6bb35 to
37c70e4
Compare
|
Is it actually the case that #136 was unnecessary since putting select columns in parentheses was possible before? |
If you mean double quotes, then yes. Quotes automatically caused it to fallthrough to this logic, making the index valid. I guess this was more of a documentation problem, but now that #136 is running in prod we'd need to rename the index if we want to revert it. @marinelli, can you verify if you started depending on this functionality, i.e. did you already create an index on an unquoted |
I think we did that. |
|
I'd prefer we revert this change and fix a few stray indexes in journey rather than pile up more code for workarounds and special cases 🤔 |
Yeah, makes sense. I'll close this and revert #136 on master and add a documentation note instead somewhere. |
My changes in #136 didn't account for legacy indices that used a workaround to get columns with non-reserved keywords by explicitly quoting them in the includes. These indices historically got generated names with the
"converted into$.This PR changes it so index column identifiers are stripped of outer quotes at construction. Indexes created by older versions still validate via a reconstructed legacy name candidate.