rebase

typos

ok but some questions/remarks have not been addressed...

Thanks for the DatetimeCache & co.

I agree having access to the logs of the task (more or less) in real-time would be very handy (as one can expect on any CI-like tool nowadays).

but please add a comment in the ArchivePostgreSQL's version of snapshot_get_heads explaining why it's (for now) a duplication of the other implementation, thanks

In D5862#150645, @aeviso wrote:

OK for the first two items, but I don't agree on the third one. The idea is to replace one of them by a direct SQL query in the near future so reworking this will be useless. I just didn't implement the query because I needed to move forward with the other stuff

thanks, the diff looks much simpler now :-)

I know I am rambling, but could it come with some testing?

use 'branches' instead of "revisions" as section in the yaml file

Mostly nitpicking comments, but I'd really prefer that:

• the cache is kept properly typed
• the cache clearing thing gets its own git revision

Add a revision to fix the annotation of InMemory's value_sanitizer

small simplification in the docstring

use get_journal_writer in test_stream

fix docstring, thx ardumont

Thanks for the split. Almost there, but a few nitpicks/comments remains:

Thanks a lot for the split.

update the copyright in __init__.py

improve commit message

rebase

rebase

rebase

rebase and adapt to current master

rebase on current master

rebase

revert the wrong fix but make it depend on D5868 instead

add tests for dates < epoch and microseconds > 0...

please take care of not including unrelated changes in your revisions/diffs

docstring, annotation and better testing

There is still the if revision (RevisonEntry.parents) that needs a fix, otherwise ok.

I still disagree with some parts of this (the blob vs. blobid naming stuff, and more importantly the whole data dict being given as argument of insert_relation() when only one element of this dict is actually needed), but meh.

There are several comments that have not been addressed

I won't discuss more the file format for the tests, it's not that critical, but please remove the implementation of __hash__ methods if they are actually not needed, and simplify the code the the __eq__ as suggested.

Overall ok, but a few things:

• I don't think we need the __hash__ methods at all (unless you want to rewrite the __eq__ ones using hash comparison)
• we usually don't put print statement in tests
• I'd prefer the more explicit file format (see inline comments), then it does not really need to be documented in the README file (but the comment in the code needs to be removed)

I may have other comments to make, but definitely, I'd really prefer having 3 diffs here also...

Actually, I would have much preferred this diff to be 4 diffs also ;-)

rebase