oops (thx anlambert)

more fix

missed lists

rebase

thx olasd

closed by 3d9f69444aeba903f701486d12978ea887a0a8f8

another fix for cassandra

rebased on D3218

adapt according to ardumont's comments and (maybe) fix cassandra tests

In D3218#78437, @ardumont wrote:

My understanding of the current log failure is that the cassandra returns you
"parents" as a list instead of tuple. This then is captured by the validation
proxy.

probably, the pb of running tests with 'not cassandra'...

yeah ;)

Jsyk, in D3211 (which was missing the rest of your current diff), I took the
party of keeping the list type in the conversion function (you changed it to
Tuple here) cassandra.converters.revision_from_db and force the parents to be
a tuple within that function. This, to avoid having to change the
cassandra.storage implementations. As you took the other path, you need to
check the revision_get and other revision functions which builds list for that field ;)

In D3218#78434, @ardumont wrote:

Build has FAILED

My understanding of the current log failure is that the cassandra returns you
"parents" as a list instead of tuple. This then is captured by the validation
proxy.

add a word in the commit message about test_api_client_dicts.py removal

Adapt the code to swh.model 0.3

ok but why move the import statement in the method?

rebase + rephrase the commit message

In D3203#78052, @vlorentz wrote:

This also kills test_write_replay.py file since it does not test anything
more than what is currently tested in test_replay.py.

Are you sure? replay.py seems to have only fixed examples, while test_write_replay.py uses anything hypothesis generates

for other systems (non debian) looks a bit weird to specify he expected location for cassandra but not for postgresql.

In D3179#77530, @ardumont wrote:

I don't recall the flag needed to exclude the cassandra tests from the tox or pytest run.
That'd be great to add it in there somewhere i guess.

The commit message should at least be amended, since it does not make all model objects immutable, but only fix a couple of model classes.

@olasd any reason not to do the step 1. ? Wether or not we "fix" the remaining hashability issue by overwriting the __hash__ method, I see no harm and it makes the model (IMHO) cleaner.

Conceptually, both Snapshot.branches and Directory.entries are mappings, not arrays of objects: snapshots and directories must not contain entries with repeated names (which is something that's currently enforced for snapshots, but not for directories).

In D3177#77360, @vlorentz wrote:

We wouldn't need that if pytest had an equivalent for assertCountEqual :/

improve the commit message

In D3177#77354, @douardda wrote:

In D3177#77353, @douardda wrote:

In D3177#77343, @vlorentz wrote:

What is the motivation for this?

Be able to create sets of revisions.

Also, having frozen objects that are modifiable seems a bit fuzzy to me, to say the least.

In D3177#77353, @douardda wrote:

In D3177#77343, @vlorentz wrote:

What is the motivation for this?

Be able to create sets of revisions.

In D3177#77343, @vlorentz wrote:

What is the motivation for this?

Forgot to update hypothesis strategies also.

Bump dep on attrs_strict >= 0.0.7

typo

bump dep on swh-model to 0.2

use ModelType instead of T for type annotation

typos, rename a couple of tests, and better comments

properly annotate BaseModel.anonymize()

Typos + comments/docstrings + hash on the fullname in Person.anonymize()

Replaced by D3171 + D3172

In D3171#77185, @vlorentz wrote:

Shouldn't we make anonymized objects error when their compute_hash() method is called?

Replaced by D3171 + D3172