builtin.jq: whilst

[pkoppstein]

whilst is like while but more conventional.

[nicowilliams]

We need to define a code style for jq. I've become very happy with 4-space indentation except for 2 spaces for |, like this:

    .
  | stuff

I think until we decide on a style for jq code, maybe it'd be best to not restyle existing jq code.

[pkoppstein]

I was just trying to ensure while/whilst/until were consistently formatted, admittedly in the 2-space style -- two spaces saves space-time, no? I thought that was important for builtin.jq ...

[nicowilliams]

Changing style of surrounding code to be more consistent is something I like to do -though it can be irritating because it clutters diffs and complicates use of git blame-, but in this case I'm not sure what style we ought to adopt, and my own jq-coding style has change over time, so for src/builtin.jq I would prefer we don't gratuitously re-style until we decide on a style. Conversely, I'm not asking you to adopt a style for jq code for now either -- if eventually we have to restyle all of src/builtin.jq, so be it.

(For C code in jq the style is fairly consistent already, and very much in @stedolan's C style.)

[pkoppstein]

@nicowilliams - In case you haven't noticed, @itchyny's builtin.jq in gojq uses the two-space style :-)

https://github.com/itchyny/gojq/blob/main/builtin.jq

[nicowilliams]

@pkoppstein no, I've not noticed. It's on my list to eventually go take a look at fq, jqjq, gojq, rq, and others, but my time is limited. What they do for style is interesting, but what they do for semantics is much more interesting as we want jq-alikes to interoperate. As for style, like I said, my own jq style has evolved and I think we should discuss that separately and not gratuitously restyle src/builtin.jq for the time being even though I otherwise would prefer that we do restyle as we go.

[itchyny]

Any reference to convince this naming is adopted somewhere else?

[nicowilliams]

whilst is like while but more conventional.

More conventional? I think I made a mistake making until/2 so different from while/2, but whilst/2 isn't more like until/2, so I don't quite understand the motivation. Can you show us a problem that's easier to solve with whilst/2 than with while/2?

[pkoppstein]

@itchyny wrote:

Any reference to convince this naming is adopted somewhere else?

Who knows? I couldn't think of a better name, and the functionality of the proposed whilst is very similar to the existing while, as the word "whilst" is similar to "while". More importantly:

In some Midlands dialects, there has been a distinct difference - "whilst" meaning "during the time when" and "while" meaning "until, up to point when", e.g. "Look after this whilst I'm gone" or "Look after this while I come back". Sounds wrong to the purist, but in Rutland they would know what you meant.

In other words, this distinction which can be drawn in English parallels the distinction between the proposed and existing jq defs.

(I was toying with emit_while but that would probably cause even more dissatisfaction because it suggests a closer affinity to jq's while than it has.)

[pkoppstein]

@nicowilliams wrote:

Can you show us a problem ...

Believe me, whilst/2 was born of necessity. On the web:

• https://rosettacode.org/wiki/Abelian_sandpile_model/Identity#jq
• https://rosettacode.org/wiki/Numbers_k_such_that_the_last_letter_of_k_is_the_same_as_the_first_letter_of_k%2B1#jq

By saying "whilst is like while but more conventional", I meant that this new def is somewhat similar to jq's existing while but is closer to traditional while loops such as C's:

https://www.tutorialspoint.com/cprogramming/c_while_loop.htm

Perhaps those familiar with such "while" loops would prefer that the "while" in builtin.jq behaved like that, but that's water under the bridge.

[wader]

We need to define a code style for jq. I've become very happy with 4-space indentation except for 2 spaces for |, like this:

For larger jq source files i've settle on a haskell/elm-ish style, 2 space indent and then explicit parentheses if some pipelines ends up multiline to make them align nicely and looks similar to array and objects literals. Also for some reason started using underscore to prefix for local functions. Can have a look at jqjq if you want an example.

Also wish there was a jqfmt (and vet/lint) :)

[itchyny]

Honestly feels kind of a literature that non-native speakers cannot tell the difference (or even worse cannot spell correctly).

[nicowilliams]

Honestly feels kind of a literature that non-native speakers cannot tell the difference (or even worse cannot spell correctly).

The main problem is that I get no real sense of the semantic difference between while/2 and whilst/2 from the two functions' names. English is my third language, and that might be why, but at any rate, I'm asking @pkoppstein to explain how this new builtin is useful.

[pkoppstein]

@nicowilliams wrote:

explain how this new builtin is useful.

References to code examples were already provided - see "rosettacode" refs above.
But the important point is that the utility is similar to C's while.

As for naming - I don't see a huge problem, as manual.yml makes the distinction quite clear.

As for mixing them up - well, we already have that problem because jq's while has a different logic than C's.

If you don't like whilst, I doubt you'll like any of the other alternatives I've considered - whilst_do; while_recurse; if_recurse; cond; ...

[@itchyny - Since the name while has already been taken, what alternatives would you suggest?]

[pkoppstein]

PR #2767 adds abs so this ER could be closed but for one point - the status of $number|length.

Until recently (that is, prior to 21 June 2023), v1.6 of manual.yml was silent about $number|length (*), and so
if one had wanted to, the v1.7 version could easily have continued to be silent.

Since it would still be possible to deprecate or discourage length on numbers, for now I'll leave this issue open for further comment.

It might also be worth reflecting on the wisdom of being cautious about retroactively adding specification detail to prior versions of manual.yml, at least when there's the probability that its omission was intentional (the probability here being related to doubts about the superabundance of polymorphism).

---

(*)

dd5ce98 docs/content/manual/v1.6/manual.yml (Alex Jurkiewicz 2023-06-21 19:21:39 +1000 750) - The length of a number is its absolute value.