Fixes to the problem of NULL equality in the selector fields.#26
Fixes to the problem of NULL equality in the selector fields.#26seamusabshere merged 4 commits intoseamusabshere:masterfrom pnomolos:null_equality
Conversation
|
So it'll add some extra work, but if we pull the table definition for the table being inserted to, it's possible to use rowid in conjunction with something like this StackOverflow comment to be able to update records. The only place where this wouldn't work is in the case of a table with no primary/unique key and a named column that overrides the internal rowid column (which seems like a fairly edge case to me). |
|
have you run the whole test suite? do the performance tests still pass? :) |
|
(also: please add this to the CHANGELOG and add yourself as an author in the README) |
|
Shall I bump the version number as well, to 2.0.2? |
|
Or I guess maybe 2.1.0 since this technically introduces some backwards compatibility? |
|
i would say a bump patch - this is a "bugfix," albeit a VERY important one agree? |
|
I guess so, the cases where someone would be depending on the previous (incorrect) behaviour don't make any sense to me :) |
|
agree - 2.0.2 - plz make sure you've got the fix for #24 in there too |
|
BTW, the one performance spec that now appears to fail is the "Upsert can be speeded up with upserting compared to activerecord-import is faster than faking upserts with activerecord-import" spec, but only on MySQL. |
CHANGELOG
Outdated
There was a problem hiding this comment.
it should be like
2.02
bug fixes
- yourfix
- the fix for 24
Fixes to the problem of NULL equality in the selector fields.
|
released as 2.0.2 - thank you! |
Fixes for issue #25. Doesn't fix Sqlite (yet).
One note - the "uses nil selectors (another way of checking)" spec is now failing in Postgres - the Date value being returned is in a different format than expected. I'm not sure if this is a problem or not :)