Wed, Aug 15
I just realized what this error message means: Some test (at the moment I don't know which) creates an Item, but this Item ends in code that expects a Lexeme.
Tue, Aug 14
[…] Thiemo already did analysis/review on the library (correct me if I'm wrong).
Mon, Aug 13
not that anyone cares if an integer is passed as a float […]
From the header of Assert.php:
Note that assertions evaluate expressions and add function calls, so using assertions may have a negative impact on performance when used in performance hotspots. The idea of this class is to have a neat tool for assertions if and when they are needed. It is not recommended to place assertions all over the code indiscriminately.
I reviewed patch set 6 of the patch https://gerrit.wikimedia.org/r/451611, and can confirm that it successfully replaces the icon as requested. But the patch does have two issues, both visible in this screenshot:
- The new icon is not centered on it's button.
- The slider UI below is partly shifted and appears broken with this patch. The vertical yellow bar belongs to the left side of the diff, but is shifted to the right. The blue bar does not connect to the circle any more.
Fri, Aug 10
As a user reported in T198711 the mistake in the code responsible for this error also created a user-facing issue. We already resolved this via https://gerrit.wikimedia.org/r/448565, and backported to the wmf/1.32.0-wmf.14 branch via https://gerrit.wikimedia.org/r/448801. We did not bothered backporting to the wmf.13 branch since Commons was on wmf.14 already at this point in time.
Thu, Aug 9
Sun, Aug 5
Thu, Aug 2
Tue, Jul 31
This is achieved through the use of group parameter […]
It looks like this feature was added in 2010 via T24265: Allow references to be listed with letters. Notice the discussion that followed after it was implemented. I'm working with <ref>s since they have been introduced, and are very familiar with the syntax and it's limitations. I'm very surprised this misuse of the existing group="…" parameter for styling was accepted.
Mon, Jul 30
Sat, Jul 28
Fri, Jul 27
I gave this a bit more time and agree that what we are discussing here is not different from, for example, (boolean) vs. (bool). However, what I don't want us to do is to disallow is_null without a proper auto-fix. is_null appears about 1600 times in 900 files: https://codesearch.wmflabs.org/search/?q=is_null\(&i=1. Having to manually deal with these would be tedious and mostly pointless, given we all agreed there is not really anything wrong with is_null. I believe such a sniff must have an auto-fix that's clever enough to understand edge-cases like the one I listed above, as well as being able to replace !is_null( $foo ) with $foo !== null.
Are we really checking characters? I seem to remember something else but am not entirely sure.
Wed, Jul 25
Tue, Jul 24
This got renamed and moved to the "ready for pickup" column without further discussion – at least none I can see at the moment. As I tried to explain in T199076#4407943 I am certain the Commons user account must be created. Otherwise the user can't find his own contributions any more (after the file got deleted at the source). Being able to find my own files on
as well as
and others is an essential (if not the one critical) feature FileImporter is build for.
Sun, Jul 22
Thanks for bringing this up!
Wed, Jul 18
is_null can make code more readable in cases where an extra pair of lose parenthesis would be needed anyway to make sure the operator precedence does not conflict with the intent, or when it simply helps making the intent more clear.
I struggle to understand how either of is_null() or === null could be clearer than the other […]
Every interaction I had with @matej_suchanek so far was an absolute, joyful breeze. I totally support this.
Guys, I honestly don't think this is worth anybodies time. Yes, isset is problematic because of it's hidden semantics. But what is problematic about is_null? Is it not readable? Does it not do what it says it does? What is the gain win by actively forbidding it?
Jul 16 2018
[…] annoying is for people to -1 patches very quickly […], but refusing to appear on this Phab task […]
Jul 15 2018
Jul 13 2018
Come on. It's not even 3 weeks old.
Jul 12 2018
Personally, I don't think it makes sense to touch the string datatype in any machine-readable output. (In other words: I don't think this is worth a breaking change.) But I think it makes sense to prefer $propertyType as a variable name in code, and "property data type" in documentation. Especially the later. What can you loose if you do this?
Jul 9 2018
Legally, every edit requires the user to agree to the license. Even if the license is CC-0. Even if the edit removes something. Every edit becomes part of the CC-0 dataset, and this can only be done if every edit is compatible with CC-0.
This ticket confuses me. Auto-creating the user in a CentralAuth a.k.a. SUL environment is the right thing to do, and a technical prerequisite for other features we want (e.g. properly linking the edits to the users own [[Special:Contributions]]).
Jul 8 2018
Please change this code to not use any of the Wikibase constants, but read the incoming format as MIME type. E.g. see if the format string starts with "text/html".