Feature: Add sized_binary(combinator, size_combinator)

[dashbitco]

Feature: Add sized_binary(combinator, size_combinator)

[josevalim]

Hi @jeffrom! The idea is to do something more efficient, where we read the size and then we perform <<x::size(y)-binary>>.

[jeffrom]

Just curious: I've implemented something that i think achieves this, which basically does a traverse to get the length from the input and adds it to the context, and then a repeat_while to increment a counter, also in the context, and repeats ascii_char([0..255]) until the counter has reached the input size. Would what you have in mind be more efficient than this approach?

[jeffrom]

definitely looking forward to that!

[w-sanches]

@josevalim is this still wanted and has no one working on it? I could give it a try.

[josevalim]

Still wanted, yes. :)

[teamon]

I stumbled upon a similar (I think) case: parsing <N><...something...><N times X>

In theory, it could be represented as (name pretty random):

def times_from_count(count_combinator, next_combinator, repeated_combinator)

and used like this:

times_from_prefix(integer(), string("---"), ascii_char(?a..?c))

and it would parse the following examples:

"2---a"             -> error, expected 2 chars after "---", got 1
"2---ab"            -> ok("ab"), "" left
"2---abc"           -> ok("ab"), "c" left
"10---abcabcabca"   -> ok("abcabcabca"), "" left

Is this anyhow similar to what sized_binary would do?
Is there any way to implement it with current nimble_parsec?

I could try to implement it with a bit of guidance of how it all works inside 🙃

[josevalim]

I have just pushed the ability to return rest from post_traverse which would allow someone to implement this by hand, so this is no longer relevant. :)