new op.f() construct not rendered correctly within an Index

[sqlalchemy-bot]

Migrated issue, originally created by Iuri de Silvio (@iurisilvio)

I know it is difficult to fix these bogus indexes, but they are failing my migration.

My upgrade drop my index and create a new one, maybe because names didn't match:

    op.create_index('op.f('ix_residencial_Categoria TV')', 'residencial', ['Categoria TV'], unique=False)
    op.drop_index('ix_residencial_desembolso_tv', table_name='residencial')

The downgrade just revert the change, with the same error. You probably have to escape the quotes or use double quotes instead.

It happens with 0.6.4. The 0.6.3 generates bogus indexes but at least doesn't break with SyntaxError.

[sqlalchemy-bot]

Michael Bayer (@zzzeek) wrote:

i dont understand the context. what is:

op.create_index('op.f('ix_residencial_Categoria TV')', 'residencial', ['Categoria TV'], unique=False)

with the quotes around "op.f()"? autogenerate created that? or you typed that (and if so, why?) ?

[sqlalchemy-bot]

Iuri de Silvio (@iurisilvio) wrote:

Yes, the line was autogenerated. I didn't changed the generated script.

residencial is my table and Categoria TV is the column to be indexed
with Column(..., index=True).

[sqlalchemy-bot]

Michael Bayer (@zzzeek) wrote:

Fixed another bug regarding naming conventions, continuing
from 🎫183, where add_index()
drop_index() directives would not correctly render the f()
construct when the index contained a convention-driven name.
fixes #194 re: #183

→ a9fa7d9

[sqlalchemy-bot]

Changes by Michael Bayer (@zzzeek):

• changed status to closed

[sqlalchemy-bot]

Michael Bayer (@zzzeek) wrote:

can you try out git master please and make sure it's resolved on your end, thanks

[sqlalchemy-bot]

Iuri de Silvio (@iurisilvio) wrote:

I tested it. Works for me. Thanks!

[sqlalchemy-bot]

Michael Bayer (@zzzeek) wrote:

thanks

[sqlalchemy-bot]

Chris Reeves (@krak3n) wrote:

Any news on a release that would include this fix, currently alembic is unusable for me considering I can't put indexes on fields.

[sqlalchemy-bot]

Michael Bayer (@zzzeek) wrote:

from what I can tell, nothing i did in the fix here should have an impact on bogus indexes being created, the issue I could see is that the "op.f()" directive, which is only there if you're using SQLA 0.9.4 or so with the latest alembic, as well as using anonymously-named indexes (e.g. index=True) which will make use of naming_convention.

so to that extent I'm not exactly sure why luri reported that it all works now, the "op.f()" should render correctly but rendering happens long after we've determined what's to be added and removed.

If you are seeing index add/removes that are incorrect, we've already had three releases each of which resolved ever more issues with this new feature and I have no test cases left that illustrate this issue occurring. So would need positive confirmation that if you run Alembic on master the issue of bogus indexes doesn't exist for you, I'm not seeing how that's possible at the moment.

[sqlalchemy-bot]

Michael Bayer (@zzzeek) wrote:

oh and additionally that you are on 0.6.4 and are seeing the issue on 0.6.4, thanks.

[sqlalchemy-bot]

Changes by Michael Bayer (@zzzeek):

• changed title from "Bogus create/remove indexes raising SyntaxError" to "new op.f() construct not rendered correctly within"

[sqlalchemy-bot]

Michael Bayer (@zzzeek) wrote:

also @iurisilvio the original error you report seems like either there is a case convention mismatch going on (e.g. mixed case names in your model are sent without quoting and therefore come back as all lower case), or you just changed the name of the column from "desembolso_tv" to "Categoria TV", is that the case?

[sqlalchemy-bot]

Iuri de Silvio (@iurisilvio) wrote:

Just to confirm, I never had this op.f() issue again, your fix works nice.

I know we still have these bogus add/drop indexes, but I can handle that without it and other issues already address this problem.

[sqlalchemy-bot]

Michael Bayer (@zzzeek) wrote:

I really need to know what case causes more index add/drops. There are dozens that I have fixed, covering all kinds of edge cases. What more are there?