Nothing is new_

[wardi]

Targets for renaming:

• templates/package/new_package_form.html Two-stage dataset create + introduce some functional tests for new_tests #1659
• new_authz.py rename new_authz.py to authz.py #1995
• old tests become tests/legacy [#1753] Rename tests -> tests_legacy. #2221
• new_tests become tests rename new_tests -> tests #2349

[tobes]

:)

All I would say based on experience is that git does not like file renames when they are ambiguous. eg. if we change test -> tests_legacy and then test_new -> tests then this will create lots of noise between branches.

new_authz.py can probably become authz.py as that file has not existed for some time although this may risk breaking badly written extensions. maybe it should just move to lib.authz.py instead.

Templates:
As you know these will break extensions. If I had my way I would just go straight to your final version 3 in #1659 the templates are too complicated by far so why not just rationalize them.

Also delete every legacy template and every failing test caused by this (yes you would loose test coverage but test coverage of unused templates/workflows is sort of pointless) - @seanh will never agree to this I suspect.

[seanh]

I'd be happy to delete the legacy templates and their tests, actually, as I doubt the tests are very helpful. I'd be even happier to see new frontend tests though!

I agree with all the renaming if it can be done practically

[tobes]

Do we also bite the bullet and change templates/package/package_form.html to templates/dataset/dataset_form.html that would be the real goal :)

[joetsoi]

+1, just a warning that we might break extensions that have imported new_tests.

e: https://github.com/ckan/ckan/blob/master/ckan/ckan_nose_plugin.py#L20

when renaming new tests, this needs to be fixed as well.

[wardi]

@tobes I don't think that's likely to happen until the next complete template rewrite (same for action api package methods) but I do support it

[wardi]

@tobes lib.authz is a good suggestion, it seems untidy to have it at the root and that would match better with lib.search etc.

But it's always seemed funny to me that things that call other parts of ckan are under 'lib' though. 'lib' to me says "separate library we've pulled in to our source tree". The authorization is tied deeply with the rest of ckan.

[amercader]

Do we want to bite the bullet on some of these? I personally would like to see the new_authz and new_tests renamed before 2.4, but if it causes too much bother we can defer.

As per deleting the old templates, see #1772

[wardi]

Yes, I would like to try to knock some of these off before they are even more entrenched

[wardi]

Thank you @vitorbaptista @amercader @davidread @joetsoi !!!

[davidread]

👍