tpolecat/atto

friendly little parsers • latest version is 0.7.0

Ryan Zeigler

@rzeigler

exprParser appears left recursive. That is usually a danger sign.

Adelbert Chang

@adelbertc

hmmmmm

good point

Adelbert Chang

@adelbertc

time to read https://en.wikipedia.org/wiki/Left_recursion

tpataky

@tpataky

Hi, I'm having a bit of trouble with reporting back the precise cause of parsing failures. I've got something like this:

  val expr = ifExpr | letExpr | ... | identifier | literal

If my ifExpr or letExpr parser successfully parses the 'if' or 'let' token but fails later, I'd like to report this as the reason of the failure. Is there a way to achieve this?

tpataky

@tpataky

No? Is there no combinator to limit backtracking? Can I not achieve this even if I restructure my parsers?

Jakub Kozłowski

@kubukoz

Hey, does anyone know of a bridge between atto and scodec?

Rob Norris

@tpolecat

Nope.

Jakub Kozłowski

@kubukoz

I went with something as simple as using the Parser as a Decoder (that part is quite trivial to implement) and passing a "render" function A => String, so I guess there's not really much room for improvement

Christopher Davenport

@ChristopherDavenport

@tpolecat Is there a way to tell what the excess elements passed to a ParseResult is? i.e. is there a way to tell if I feed in more string which string is the overrrun that went from Partial to Complete?

Rob Norris

@tpolecat

Like, you want to feed in character by character and stop when it's complete?

Christopher Davenport

@ChristopherDavenport

Basically trying to reclaim the excess bits of string to return to the stream that way I can keep streaming and change the parser after the parser is complete.

Basically I grab exactly 1 line of headers, and then want whatever is left over to be turned over to a different parser that will read the rest as a different format. But I can’t use atto for the whole structure because it needs to stream.

I can do character by character, since its likely very feasible, but thats probably not very efficient.

Christopher Davenport

@ChristopherDavenport

I think I lose arbitrary characters in the streaming implementation to the ~parser~ ParseResult, and since I can’t recover them that means my next inputs won’t be valid.

Christopher Davenport

@ChristopherDavenport

Ignore my stupidity.

I missed the positional drop involved in the done constructor. So that its no longer input but is only the excess. My apologies for the misplaced alarm.

Rob Norris

@tpolecat

:+1:

Oleg Pyzhcov

@oleg-py

Hello there. Is it possible to force a parser to consume all input or fail instead of giving "leftovers" in Done?

Rob Norris

@tpolecat

You can do p <~ endOfInput and it will fail if there is more data.

Christopher Davenport

@ChristopherDavenport

Am I incorrect in my understanding that technically the toString implementation of Parsers breaks the Functor and Monad laws? As .map(f).map(g) is different than .map(f.andThen(g)

It’s toString, so its all cheating, but wanted to make sure we were cheating on purpose

Rob Norris

@tpolecat

yes

toString is a side-effect

Christopher Davenport

@ChristopherDavenport

Cool, works for me. I’m currently working on a general representation of Atto over abstract inputs, and was trying to understand whythat was in place.

Rob Norris

@tpolecat

Aha, ok. I would look at Oscar's talk from SBTB last year and build something based on that.

Atto is really weird.

Christopher Davenport

@ChristopherDavenport

Wish you told me that, like 3 hours ago… :laughing:

I’ve got everything working all the way up to combinators now. :sweat_smile:

Rob Norris

@tpolecat

:-)

Soren

@srnb_gitlab

This is probably asked a lot but I noticed atto-fs2 allows me to stream a String into a parser

So I was wondering

I want to get a Stream[F, Expr] out, i.e. parse expression by expression

But often at times I'll have expressions that have sub-expressions

I was wondering if I could lazily parse those sub-expressions?

I was taking a look over your scalebythebay talk about functional programming with fixed point types

And I feel like this is similar

Where in that, ProfF definitely had sub-IDs that you referenced later, Expr maybe has sub-exprs that I want to get later

trait Expr
case class FunctionF[F[_]](name: String, subExprs: Stream[F, Expr])
case class Function(name: String, subExprs: Vector[Expr]) extends Expr

I'm wondering how feasible something like this is? I'm recalling your ProfF example from memory and changing it to Stream and Vector, but I may have gotten something else wrong

Soren

@srnb_gitlab

Now, there's other Exprs than Function, and I'm also not sure I ever want to concretely keep a Vector[Expr]

Soren

@srnb_gitlab

How do I make a parser not consume (lookAhead)? I.e. string("//") ~> manyUntil(anyChar, lineTerminator.lookAhead) for a comment?

Alex Antonov

@antonoal

Hi. Is that right that when I've got Done("some non-empty string", ...) and I call .either on it I'm getting Right rather than Left with the input that couldn't hasn't been parsed?

Rob Norris

@tpolecat

Leftover input is ok. If you want to disallow it you can say p <~ endOfInput which will then fail if there's anything left over.

Alex Antonov

@antonoal

ok, will give it a try

thanks!

Rob Norris

@tpolecat

Or phrase(p) if you prefer.

Bartosz Szmit

@bszmit

Hi ;)
Is it possible to implement arithmetic grammar WITH substitution and division in atto?
Grammar like this one (I know it's left recursive)

<Exp> ::= <Exp> + <Term> | <Exp> - <Term> | <Term>
<Term> ::= <Term> * <Factor> | <Term> / <Factor> | <Factor>
<Factor> ::= int | ( <Exp> )

So far I have only seen grammars without substitution and division. They require left-associativity and I guess it's a reason they are often left out in examples.

Rob Norris

@tpolecat

No, not in that form. Recursive grammars end up being tricky.

Partly because they're just tricky, partly because atto is derived from Haskell code so it's stricter than it should be in some places, partly because of bugs.

Rob Norris

@tpolecat

A pattern that works iirc is, expr = listof(addend + addened) and then addend = listof(factor * factor) and factor = int | (expr). So it only recurses in the rightmost position. It's been a while, sorry. There should really be an example.

Where communities thrive

People

Repo info

Activity