Good question which I cannot answer :-)

The idea is that you could use whitespace patterns ('Squeak *alpha') interchangeably with more powerful regular expressions ('Squeak \d+\.\d+alpha' asRegex) without needing to define separate messages wherever patterns are used. For example, we would need implement both #assert:matches: *and* #assert:matchesRegex: (TestCase), both #classesIn: *and* #classesInRegex: (provided that we wanted this feature) etc., each pair with exactly the same implementations that only deviate in terms of sending #match: vs. #matches:.
(For a practical example, you could take a look at all these methods from SqueakByExample that contain 'match' in their name<https://github.com/codeZeilen/SqueakByExample-english/tree/master/SmalltalkSources/SBE-Extract.package/SBEWorldHelper.class/instance> and which all use #matches:. This project also includes a String >> #matches: extension so it is much easier to feed these morph query methods either with regexs or with simple patterns where regexs are unnecessary complicated.)

I agree that it might be confusing if you forget the #asRegex. However, I think this is a common problem in a language with duck typing and very-late binding. On the contrary, one could also argue that it might be confusing that in some situations you need to send #match:, but in other situations #matches: instead, as they basically do the same.

---

Hm, this discussion reminds me a bit of the recent debate about Symbol >> #numArgs/#value:value:. There we found that instead of handling symbols like blocks, we could also write an explicit converter (#asBlock). How would you think about a similar converter in this case? :-)
Something like: 'Squeak *alpha' patternAsRegex
'ab*' asRegex matches: 'a'. "true"
'ab*' asRegex matches: 'abb'. "true"
'ab*' patternAsRegex matches: 'a'. "false"
'ab*' patternAsRegex matches: 'abc'. "true"

Pro: Actually a matcher - you don't have a string that partially behaves like a matcher and follows the same protocol, but an actual matcher. So you could also say 'ab*' patternAsRegex matchesIn:collect: ...
Contra: Actually a matcher - this can be a significant drawback in terms of performance. For a simple '*xyz*' match, the regex equivalent was more than 100x slower than the whitespace version.
A third variant would be an extra WhitespaceMatcher. But do we really need such a class?

Best,
Christoph
________________________________
Von: Squeak-dev <squeak-dev-bounces at lists.squeakfoundation.org> im Auftrag von Jakob Reschke <forums.jakob at resfarm.de>
Gesendet: Freitag, 10. April 2020 11:36:37
An: The general-purpose Squeak developers list
Betreff: Re: [squeak-dev] The Inbox: Collections-ct.885.mcz

The patterns matched by match: are not regular expressions. Could this
intermixing of vocabulary lead to confusion or even bugs (when someone
is missing a test for their matching use case and forgot the asRegex)?

Am Do., 9. Apr. 2020 um 14:39 Uhr schrieb Thiede, Christoph
<Christoph.Thiede at student.hpi.uni-potsdam.de>:
>
> See also SUnit-ct.127 (#assert:matches:) or Tests-ct.429 (#assertText:) in the Inbox which could benefit from this extension.
>
>
> I would also welcome to completely rename #match: into #matches: for the reasons described below, how would you think about this?
>
> Or should String >> #matches: rather be an extension method from the Regex-Core package?
>
>
> Best,
> Christoph
> ________________________________
> Von: Squeak-dev <squeak-dev-bounces at lists.squeakfoundation.org> im Auftrag von commits at source.squeak.org <commits at source.squeak.org>
> Gesendet: Donnerstag, 9. April 2020 14:35:36
> An: squeak-dev at lists.squeakfoundation.org
> Betreff: [squeak-dev] The Inbox: Collections-ct.885.mcz
>
> Christoph Thiede uploaded a new version of Collections to project The Inbox:
> http://source.squeak.org/inbox/Collections-ct.885.mcz
>
> ==================== Summary ====================
>
> Name: Collections-ct.885
> Author: ct
> Time: 9 April 2020, 2:35:32.800949 pm
> UUID: ddccd842-af02-a448-91f7-7556c8c7bf43
> Ancestors: Collections-dtl.884
>
> Proposal for discussion: Add String >> #matches: for polymorphy with RxMatcher.
>
> This allows you to pass patterns strings whereever a RxMatcher is expected. Furthermore, IMHO the name #matches: gives you a better intuition for a test selector than #match: does.
>
> | password patterns |
> patterns := { '.{8,}' asRegex. '*password*'. '.*\d.*' asRegex }.
> password := Project uiManager requestPassword: 'Enter a fake password'.
> self assert: (patterns allSatisfy: [:pattern | pattern matches: password]) description: 'Your password is not safe enough'.
>
> =============== Diff against Collections-dtl.884 ===============
>
> Item was added:
> + ----- Method: String>>matches: (in category 'comparing') -----
> + matches: text
> +        "For polymorphy with RxMatcher."
> +
> +        ^ self match: text!
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20200411/3ae7d435/attachment.html>