4
$\begingroup$

I recently came across Andrea Mori's question Explicit Casselman theory: reference needed. The comments 1 2 make reference to matrices rendering poorly, and then being fixed, and I was curious what had changed, so I checked the version history. The relevant version, version 5, seems to have no content. I don't know how to link to the diff (and Meta.SE suggests it's not possible), but, if you click the drop down on the relevant revision on the version history, then you'll notice that … nothing happens, or, rather, the diff appears to be empty. How did an edit that must have had content just disappear?

$\endgroup$
9
  • 1
    $\begingroup$ Nothing mathematical it seems. Here's the revision comment: "the matrices weren't displayed rightly, so I placed backticks around the equations." Note that the asterisk inside TeX used to cause problems, and the hacky fix was to use backticks to somehow escape the Markdown interpretation of *. (The better way around this was to use \ast) But this was fixed, and it seems all the hacky backticks removed. My guess is that this post got eaten somehow. $\endgroup$
    – David Roberts Mod
    Oct 23, 2023 at 21:29
  • 1
    $\begingroup$ I have a vague memory that possibly there was a problem with some posts because of this backend change/automated edit, and and it was able to be fixed with an edit that made any change at all, as you can see in the history. But still just surmising here. $\endgroup$
    – David Roberts Mod
    Oct 23, 2023 at 21:32
  • 1
    $\begingroup$ As far as I can tell, the two revisions are the same, here are some SEDE queries: list of revisions and comparison of the two revisions. Wayback Machine probably does not help here either - all snapshots are after that edit. $\endgroup$ Oct 24, 2023 at 9:23
  • 1
    $\begingroup$ I thought at first whether the edit might have been empty at the time - but I think that in such cases the edit summary simply says Edit removed during grace period Actually, I have just tested this in the sandbox: meta.mathoverflow.net/posts/57/revisions and meta.mathoverflow.net/revisions/57/25 $\endgroup$ Oct 24, 2023 at 9:24
  • 1
    $\begingroup$ @DavidRoberts and $\endgroup$
    – LSpice
    Oct 24, 2023 at 18:27
  • 1
    $\begingroup$ @MartinSleziak, thanks! My suspicion was that whatever backtick-related kludge was needed might not only have been fixed on a per-post basis, but somehow removed from the edit history entirely. Is that possible? $\endgroup$
    – LSpice
    Oct 24, 2023 at 18:28
  • 1
    $\begingroup$ This answer is a bit similar - it has an edit which is empty (but the edit summary says "Fixed LaTeX") and at one point (in another edit) \* was changed to \ast. One can find various related SEDE queries here in chat. $\endgroup$ Oct 28, 2023 at 19:04
  • 1
    $\begingroup$ @MartinSleziak, re, thanks! Without SE staff coming by to give us a peek into the internals, that seems to be as close to an answer as we're going to get. If you would post it as an answer, then I would be happy to accept it. $\endgroup$
    – LSpice
    Oct 28, 2023 at 19:10
  • 1
    $\begingroup$ As far as I can tell, both edits were before MO joined the Stack Exchange network. So I am not sure who actually is in a better position to say something - the people who were running the site at the time, or the people who know details of the software as it runs now. (Which reminds me - did MathOverflow, Inc. release publicly some data dump at the time of joining the SE network?) $\endgroup$ Oct 28, 2023 at 19:31

1 Answer 1

2
$\begingroup$

A few comments which might be useful for somebody who could be in the mood to try to investigate this a bit more - I am basically summarizing the things already mentioned in comments or in chat.

One can try to find some other similar edits - but in doing so, you'll probably find that edits which appear to change nothing are relatively common. Regarding the SEDE queries that I posted here, you should take into account both the fact that I am not really that experienced with SQL and also some peculiarities in how SQL compares strings.1

  • This query finds revisions which did not change the length - but various fixes which replace some character by a different one are probably among relatively common types of edits. So it's no surprise that there are plenty of such edits.
  • This is a bit more relevant - no change in the body. SEDE returns 948 such edits.
  • One can try to find such edits by a specific user: LSpice, Martin Sleziak. (I have included these two users as examples. The query is mentioned in case you're curious whether some of your edits are of this type - simply replace the userid parameter by your id.2)
  • Many of those edits say "Edit removed during grace period" in the edit summary - one can omit such edits, you will still get 658 results.
  • Specifically to this question, one can check whether \* appears in the text of the old revision or of the new revision. This is how I found the answer mentioned above.
  • One of the comments speculated that all instances of \* have been removed - but this is not the case. SEDE returns 152 posts with \*. This includes \\* which can appear, for example, in a matrix. When searching for posts with \* but without \\*, I get 145 results. There are 180 posts where \* was edited.

I will add that both empty revisions mentioned here were done before MO joined the Stack Exchange network. (One of them in January 2013 and the other one in July 2010.


1Stack Overflow: SQL - string comparison ignores space

2Meta Stack Exchange: What is my user ID?

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .