fn:scan-left and fn:scan-right - produce accumulation of results

[dnovatchev]

fn:scan-left and fn:scan-right - produce accumulation of results

[ChristianGruen]

Related: https://docs.basex.org/wiki/Higher-Order_Functions_Module#hof:scan-left

[ChristianGruen]

PRs are always helpful.

[liamquin]

Looks like these could be defined in terms of the existing fold functions?

[ChristianGruen]

Looks like these could be defined in terms of the existing fold functions?

With a scan, the intermediate results are returned as well. While it’s possible to generate a sequence with the intermediate results with fold, it’s more convenient to have a custom function for it.

I know that Dimitre will disagree, but as hardly anyone uses fold-right in practice, a single scan function should suffice (e.g., as Kotlin has done it).

[liamquin]

Thanks for replying. Agree about fold-right! And in the XSLT 3 course i run, some people have said they'd struggled for ages with fold-left() before having an example and working through it. But i am not sure that they would find scan() any easier, and having two functions to learn instead of one is a burden sometimes. I’d vote for less-is-more on this one.

[ChristianGruen]

Thanks for replying. Agree about fold-right! And in the XSLT 3 course i run, some people have said they'd struggled for ages with fold-left() before having an example and working through it. But i am not sure that they would find scan() any easier, and having two functions to learn instead of one is a burden sometimes. I’d vote for less-is-more on this one.

Yes, learning folds may take a while. Until today, I don’t understand how our spec tries to explain how they work (and I wish we would revise the text; related: #864 (comment)). Instead, I learned folding by looking at other languages (or books ;).

My feeling is that once you’ve understood folds, scans are easy. Maybe we can find a better name, though.

[michaelhkay]

I have certainly found the need for something equivalent to scan-left. I agree that it's very difficult to explain these functions to users (and that's especially true for fold-right, which I still struggle with myself) and we should certainly try to do a better job on explanation (especially by example), but I don't think that's an absolute reason for leaving them out.

[dnovatchev]

Here is more about the use-cases for the scan functions: https://www.reddit.com/r/haskell/comments/s737lq/scan/

And especially here: https://www.cs.cmu.edu/~guyb/papers/Ble93.pdf

[dnovatchev]

As there seem to be no more comments, I will proceed to creating a PR.

Any objections?

[dnovatchev]

Submitted the PR, can be viewed here now: #957