i think maybe ive started down a bad / wrong path ….. if anyone has the time & all round lovelyness to make sure im not going completely off the reservation ….

i want to be able to javascript src code a text file and pull out certain parts. specifically i want to try and captue “test name” from the common it('xxxx' and describe("xxxx" style of test definitions.

so i figred parsers were the right way to go?

however because i want to ignore an awful lot of the inbetweeen parts, i seem to be running up against a lot of friction with the approach

im hoping someone can just sanity check me and let me know if there is a vastly simpler approach or just validate my thinking that parsers are probably the correct solution, and stick with it you will get there sort of thing ? :slightly_smiling_face:

thanks in advance

Did you write your own parser for JavaScript?

no, i dont want to parse the whole of the language

eventually id like the approach to me src language agnostic

so im trying to impliment the parser as small as possible and focused only on extracting the infomation i need. right now thats just test names

How general do you need it? Would a line-by-line extraction with regular expressions be enough?

mmmm possibly….

possibly this is just over engineering, but that felt like a less robust solution ?

the parser approach would be more likely to end up simpler when dealing with nested structures etc

describe('..... it('.....) )

I think you could write a 95% solution using regexps. An alternative is to use the lexical scanner from Herman’s JavaScript compiler. I don’t think you need the full parser. Simply turn the file into a stream of tokens. When you see “describe” "(" <string> you output the string. https://docs.racket-lang.org/javascript/parse.html#%28part._lexer%29

yeah im playing about with lexers right now (after starting with parser from megaparsack)

im not sure i entirely get the fundamental difference

Between?

lexers & parsers

probably one for me to jsut google myself :slightly_smiling_face:

The lexer breaks the source into small logical units like “number”, “string” “parenthesis” etc. No attempt of finding a structure is made.

Another approach: Can you define your own version of “describe” and “it” in JavaScript. Which won’t run the tests, but simply print their names. Then run the tests with your modified describe and it.

using this ? https://docs.racket-lang.org/javascript/eval.html

i suppose that’s an option, but it sort of closes the door for other languages unless i can go a similar route of comming up with an interpriter / compiler & monkey patches for the test functions / classes etc

I was thinking of using, say, node to run the JavaScript. But you are right, that it is a JavaScript only approach.

sorra gotta run, but ill check back in later :slightly_smiling_face: really appreciate the advise / sounding board :slightly_smiling_face:

@chris613 Personally I would write the regexp-solution first. Then replace it if it turned out not to be good enough.

Can I use for/or to pick out the first element in a list for which some predicate is true? Like (for/or ([i '(1 3 5 7 8 9)]) (even? i))

but what I like is the actual value, not #t/#f.

(for/or ([i ’(1 3 5 7 8 9)]) (and (even? i) i)

Or use #:when

(for/first ([i '(1 3 5 6 9)] #:when (even? i)) i)

Nice. Reading the docs for for/or and for/first I find it hard to tell quite what the difference is between them.

for/or continues when it sees #f. for/first will return #f

So for/or returns the first non-false value and for/first returns the first value.

So comparing the two approaches above, are they equivalent in terms of the number of invocations to even??

I believe so, but you can replace (even? i) with (begin (display i) (even? i)) to check.

Yes, they are.