RF: make publish understand --since=^ and deprecate --since=

[yarikoptic]

Positioned against maint since

• previous code should still work, we will just announce (well, DeprecationWarnings are not shown! by default, heh) deprecation
• it would improve my UX while switching from publish to push (although ENH(UX): explicitly disallow empty string for --since in push #4682 would also help) so I could just use --since=^ in both

Who knows, may be in 0.15 the entire publish would be gone...

[kyleam]

I have no objections to making publish understand '^' for consistency. As for applying it to maint, all right, though I think we should avoid using "it doesn't break" (should be true for nearly all code for master too) and "improves things for me" as reasons to make non-bugfix changes to maint.

However, I really don't think we should introduce deprecations on maint unless we have a very strong reason to. In this case, I think publish should just learn to treat '^' as another spelling of ''. Continuing to support the latter has virtually no maintenance cost.

Who knows, may be in 0.15 the entire publish would be gone...

It shouldn't be. We haven't deprecated it yet, so that'd be at least 0.16.

[yarikoptic]

Re no maintenance cost - there is UX cost due to more than one way to do the same, confusion etc. I wish I just did it for 0.13, where it would have been right in place.. heh heh... Also iirc now there is datalad diff which might need the same (didn't check)

[kyleam]

there is UX cost due to more than one way to do the same, confusion etc.

Yes, ideally an interface wouldn't have multiple ways to say the same thing. But in this case the tradeoff is that it allows consistency between publish and push without making users or us have to go through a deprecation sequence.

[yarikoptic]

But in this case the tradeoff is that it allows consistency between publish and push without making users or us have to go through a deprecation sequence.

it wouldn't be "consistent" unless we also allow empty string for push... the sooner we encourage (and really weakly since that deprecation warning wouldn't be shown by default) users to adhere to consistent interface, the better IMHO.

[kyleam]

But in this case the tradeoff is that it allows consistency between publish and push without making users or us have to go through a deprecation sequence.

it wouldn't be "consistent" unless we also allow empty string for push...

To reword my statement in a less sloppy, context-dependent way: "it allows users to consistently say --since=^ across publish and push".

the sooner we encourage (and really weakly since that deprecation warning wouldn't be shown by default) users to adhere to consistent interface, the better IMHO.

I disagree. I'll leave it at that, since I'd just repeat my statements from above.

[yarikoptic]

@bpoldrack @mih @adswa , any opinion? Deprecate or not --since= in publish?

[bpoldrack]

I find @kyleam's argument convincing.

[yarikoptic]

Removed deprecation.

[codecov]

Codecov Report

Merging #4683 into maint will decrease coverage by 0.27%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##            maint    #4683      +/-   ##
==========================================
- Coverage   89.92%   89.65%   -0.28%     
==========================================
  Files         288      288              
  Lines       40158    40359     +201     
==========================================
+ Hits        36114    36184      +70     
- Misses       4044     4175     +131     

Impacted Files	Coverage Δ
datalad/distribution/publish.py	88.81% <100.00%> (ø)
datalad/distribution/tests/test_publish.py	100.00% <100.00%> (ø)
datalad/support/due.py	48.00% <0.00%> (-32.00%)	⬇️
datalad/ui/utils.py	46.87% <0.00%> (-25.00%)	⬇️
datalad/support/archive_utils_patool.py	29.87% <0.00%> (-10.39%)	⬇️
datalad/tests/test_misc.py	90.00% <0.00%> (-10.00%)	⬇️
datalad/metadata/tests/test_extract_metadata.py	91.11% <0.00%> (-8.89%)	⬇️
datalad/support/tests/test_path.py	92.50% <0.00%> (-5.00%)	⬇️
datalad/downloaders/http.py	81.85% <0.00%> (-4.82%)	⬇️
datalad/support/tests/test_due_utils.py	78.57% <0.00%> (-4.77%)	⬇️
... and 66 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update fef3c70...d243432. Read the comment docs.