Update for 0.4.1

[jorisvandenbossche]

Checklist

• Used a fork of the feedstock to propose changes
• Bumped the build number (if the version is unchanged)
• Reset the build number to 0 (if the version changed)
• Re-rendered with the latest conda-smithy (Use the phrase @conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
• Ensured the license file is being packaged.

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[jorisvandenbossche]

Ensured the license file is being packaged.

@ocefpaf did I understand your comment correctly in that the full line license_file: '{{ environ["RECIPE_DIR"] }}/LICENSE.txt' can be dropped (now the license file is in the tar.gz package), or do I still need to keep something like license_file: 'LICENSE.txt' ?

[ocefpaf]

You still have to keep license_file: 'LICENSE.txt'.

[ocefpaf]

@jorisvandenbossche you can also remove the LICENSE.txt from the feedstock.

[gboeing]

@jorisvandenbossche looks like the sha needs to be updated in meta.yml

[jorisvandenbossche]

@gboeing yep, was already doing that :-)

[jorisvandenbossche]

@ocefpaf So it is failing on the testing phase, because the created environment with fiona is broken ... :) (potentially similar problem as we have been talking about today).
Is there a way to influence how the build process creates the environment to build / test geopandas? (eg use a strict channel config?)

[ocefpaf]

We should be discussed forcing strict in conda-forge builds but I don't think we'll do that soon. You just forced me to dig deep in find the culprit. (It may take a while though...)

[ocefpaf]

@jorisvandenbossche if my findings are correct psycopg2 is the culprit b/c conda-forge is in the middle of the openssl migration and conda ends up preferring an old libgdal from defaults instead of the libgdal 2.4 from conda-forge. We can:

• remove psycopg2 for now, it is an optional dependency, and re-add it later;
• force libgdal 2.4 here. bu over-specifing our dependencies we help conda do the right thing;
• or implement the strict channel policy on the feedstocks. (Pinging @conda-forge/core here for this one.)

@msarahan the issue on the staged-recipe package you pinged me a few days ago is probably similar. Sorry but I lost the link and I forgot the name of the package.

[msarahan]

conda-forge/staged-recipes#7871

[ocefpaf]

Thanks! Looking into it now to see if it is the same issue. I'm afraid that the best solution would be enforcing strict channel in our builds.

[msarahan]

I think that's a good solution. I think conda should consider changing to strict channel priority by default for conda 4.7, but before then I'd like to try to improve the messaging around unsatisfiable dependencies. A lot of people have been confused about how strict channel priority makes some packages unavailable, even though they seem to be available by inspection.

[jorisvandenbossche]

I don't understand the conda forge pinning enough to know if forcing libgdal to 2.4 could cause problems (if it is pinned like that, why do we actually not get it? Is it possible right now to install geopandas with eg gdal 2.3?).
So I would defer to your judgement. Removing psycopg2 is fine (then also sqlalchemy). That might be a good thing anyway, as it is quite optional.

[ocefpaf]

I don't understand the conda forge pinning enough to know if forcing libgdal to 2.4 could cause problems (if it is pinned like that, why do we actually not get it?

It won't cause a lot of problems but we won't be able to install geopandas 0.4.1 with libgdal 2.3.

Is it possible right now to install geopandas with eg gdal 2.3?).

It is as long as you use the strict channel 😄

So I would defer to your judgement. Removing psycopg2 is fine (then also sqlalchemy). That might be a good thing anyway, as it is quite optional.

Let's go for this option then. Do you want to also remove sqlalchemy for consistently not provide the optional deps?

[jorisvandenbossche]

Do you want to also remove sqlalchemy for consistently not provide the optional deps?

just because the one is only useful with the other

[jorisvandenbossche]

@ocefpaf thanks a lot for figuring it out!

[ocefpaf]

@ocefpaf thanks a lot for figuring it out!

Thank you for your patience. Hopefully conda 4.7 will enforce strict channel by default and all our problems will go away. (Or we will just create new ones 😜)