KeyError: 'sha1'

[he7d3r]

from mw.api import Session
from mw.lib import reverts
session = Session("https://pt.wikipedia.org/w/api.php")
reverts.api.check_rev(session, {'revid':40679746,'pageid':69874})
reverts.api.check_rev(session, {'revid': 40679734, 'pageid': 24386})

works fine, however

reverts.api.check_rev(session, {'revid':40679748,'pageid':40677})

causes the following error:

Traceback (most recent call last):
  File "demonstrate_scorer_on_rc.py", line 79, in <module>
    reverted = 0 if reverts.api.check_rev(session, rev) is None else 1
  File "<mypath>/lib/python3.4/site-packages/mw/lib/reverts/api.py", line 40, in check_rev
    return check(session, rev_id, page_id=page_id, **kwargs)
  File "<mypath>/lib/python3.4/site-packages/mw/lib/reverts/api.py", line 115, in check
    for revert in detect(checksum_revisions, radius=radius):
  File "<mypath>/lib/python3.4/site-packages/mw/lib/reverts/functions.py", line 40, in detect
    for checksum, revision in checksum_revisions:
  File "<mypath>/lib/python3.4/site-packages/mw/lib/reverts/api.py", line 110, in <genexpr>
    ((rev['sha1'], rev) for rev in past_revs),
KeyError: 'sha1'

[halfak]

Now that's a trick. The API is not returning a sha1, but we'd like to be able to detect a revert. This means that we might just give up on detecting a revert for this revision. We might also try to go to the API to look for content so we can generate our own sha1.

[he7d3r]

Weird, the API is returning the sha1 when I make a request on Special:ApiSandbox:
https://pt.wikipedia.org/wiki/Special:ApiSandbox?uselang=en#action=query&prop=revisions&format=json&rvprop=ids|timestamp|user|size|sha1&revids=40679746

[he7d3r]

I noticed the same problem with other revisions for which there is a sha1:

• 40840140
• 40841893
• 40843260

[he7d3r]

The page 40677 contains the revision 40327912, which has an attribute "sha1hidden": "" (and a sha1 which seems to be visible to sysops only):
https://pt.wikipedia.org/w/api.php?action=query&prop=revisions&rvprop=ids|timestamp|user|size|sha1&rvlimit=20&pageids=40677
However it is far from the revision we are interested in. The error does not happen if I run

reverts.api.check_rev(session, {'revid':40679748,'pageid':40677}, radius=7)

but it happens when I increase the radius to 8:

reverts.api.check_rev(session, {'revid':40679748,'pageid':40677}, radius=8)

Using a big radius should not be a reason for failure, because 40679748 was reverted by the revision which is right after it in the history (40686393), and a radius of 2 is enough to detect that (BTW: shouldn't it radius 1 be enough too?). In other words, we have something like this:
A[sha1hidden] --> B --> C --> D --> E --> F --> G --> H --> I[reverted] --> H

[halfak]

Fixed in a293667

And updated in v0.4.11