Open jamesdbrock opened 2 years ago
:+1: from me about using the term 'commit'.
Maybe an alt branch should only be commited by explicitly calling the commit function? That would be a very breaking change.
This is a very different style of parser. There's nothing wrong with it, it's just a very different way of writing parsers, and it seems really odd to arbitrarily switch to it. It's not fundamentally more expressive, or even easier, it's just inverting the mechanism while also coming with it's own set of downsides. This is essentially the age-old "backtrack by default" argument.
“commit” is the term used everywhere in Elm parser https://package.elm-lang.org/packages/elm/parser/latest/
I want to change the term “consume” everywhere to “commit.”
The term “consume” is inherited from the early days of Parsec ( Parsec: Direct Style Monadic Parser Combinators For The Real World section 3 Restricting Lookahead ), but it is very confusing. The term “commit” much better conveys the meaning of “fail the parse right here instead of trying the alternative parse.”
In addition, we should talk explicitly about “backtracking” in the documentation. We always allow arbitrary backtracking, so the entire input stream is always held in memory.
Maybe primitive parsers should never commit (and should always backtrack, as they do now)? Maybe an alt branch should only be commited by explicitly calling the
commit
function? That would be a very breaking change.The term for moving forward in the input stream is “advance,” as in
Parsing.Combinators.advance
.