This is a parser for files conforming to the ABC Notation (version 2.1) for traditional music scores. It is written in pure Elm (version 0.18) and uses the elm-combine parser combinator library. This is slightly difficult to achieve because, as far as I can make out, there is no formal grammar published, only a long textual description of the language.
The intention is to provide a fully conformant parser for individually presented monophonic ABC tunes. The grammar has been slightly relaxed in places so that it can support an application that 'plays' either the entire tune or just a portion of the tune body. So, for example, the rules that require the 'X' header to come first and the 'T' header to come second will be dropped. Obviously it is not possible to play a tune fragment if headers representing key signature, note length and tempo are absent, so these are defaulted to sensible values if they don't exist.
To parse an ABC string you can use:
abc
|> parse
Additional facilities exist in the Music modules to manipulate the parse tree in order (for example) to transpose a tune or change its octave or tempo.
- elm-lang/core 5.0.0
- Bogdanp/elm-combine 3.1.0
- elm-community/ratio 1.1.0
- elm-community/maybe-extra 3.0.0
- elm-community/elm-list-extra 4.0.0
Sections of the spec marked as 'volatile' are treated as being non-normative. The parser attempts to treat them as liberally as possible. In any case, these sections do not affect the primary purpose which is to support the replay of simple monophonic tunes.
Slurs are not implemented in a way that I like (where you recognize the left bracket, then the slurred notes and finally the right bracket; thus balancing the brackets). This seems to be a limitation in the spec which is written in a line-oriented fashion, where lines of notes may possibly be interrupted by in-line headers. However, in the wild (and perhaps the spec - it's unclear) slurs may occur across lines. This means that all that can be easily recognised is the individual brackets '(' and ') that frame the slur and which are now disconnected from each other in any meaningful sense. This means also that the parser will reject contructions like this:
A(Bc)- cd
It is clear from the specification that ties must be associated with the left-hand note of the pair and that there may be a space before the right-hand note. i.e. this is legal:
ABc- cd
whereas this is not (and will be rejected by the parser):
ABc -cd
It is not clear from the specification whether or not chords can be tied. This parser rejects them - for example:
AB [ce]- [cg]
Bar lines are specified very badly and the parser tries to make the best of a bad job.
Only simple repeat markings are supported - for example:
|1 ABcd |
[2 efga |
Complex repeats of the following form are not supported:
[1,3 ABCd :|
[1,3,5-7 ABcd :| [2,4,8 efga :|