fix #6198: stop considering generated column definition as being its default value

[allan-simon]

Q	A
Type	bug/feature/improvement
Fixed issues	#6198

Summary

The pg_get_expr(adbin, adrelid) expression in Postgresql's internal table pg_attrdef is used to know a column definition

for most column, if this returns something, it means this column's default value. So DBAL has been considering has ALWAYS meaning it contains its default.

However for generated columns, it contains the value definition and not its default value, so we change the selectTableColumns' query to correctly set the 'default' column to null for generated columns

It will help setting correctly the column's attributes which in turn will help generate correct schema diff.

[allan-simon]

I've tested manually on Postgresql 10 (with no support at all of GENERATED column, hence the feature detection code ) , Postgresql 13, 14, 15

For automated test, I've written one, but to be honest I haven't found a documentation on how to run them

[allan-simon]

Ok I've finally been able to setup the CI locally for postgresql and correct the coding style

so now I'm pretty sure all will be green :)

[allan-simon]

@derrabus I see there's still some changes requested in this PR, but normally I've adressed everything ? Is there's something you're waiting from me ?

(I don't mean to push you, if it's just a "I didn't have time yet to check it" , no worry I'll way :blush: , it's just I'm suddenly thinking "maybe he's the one waiting for me" ^^" )

[derrabus]

No, I'm aware that the ball's in my court here. 😓

[allan-simon]

repushed (I forgot to run phpcs locally) and squashed, the previous CI was only failing on IBM DB2 test, but it seems to me it was a flaky test

[allan-simon]

@derrabus the fail seems to be a flaky test , can you relaunch them ?

[greg0ire]

So DBAL has been considering has ALWAYS meaning it contains its default

Is this sentence broken? I don't understand it.

[allan-simon]

So DBAL has been considering has ALWAYS meaning it contains its default

Is this sentence broken? I don't understand it.

yes , sorry

I meant

DBAL was interpreting "pg_get_expr(adbin, adrelid) returns something" as meaning "the value returned is the default value of the column " while it's a more generic "definition value" which can be the default value , but can also be other things (like the generated column expression )

I think the confusion was that postgresql uses the abbreviated form "pg_attrdef" and one could have interpreted that as "postgresql attribute default" whereas it actually means "postgresql attribute definition"

[allan-simon]

@derrabus friendly reminder this PR exists , tell me if you need anything from me :) I can make myself available for a pair review call if you want.

[allan-simon]

thanks a lot !