diff options
Diffstat (limited to 'notes.md')
-rw-r--r-- | notes.md | 22 |
1 files changed, 22 insertions, 0 deletions
diff --git a/notes.md b/notes.md new file mode 100644 index 0000000..9eddec5 --- /dev/null +++ b/notes.md @@ -0,0 +1,22 @@ +# Problem +When in a one-of block, it's harder to tell where an error came from. + +# Current solution +We have critical blocks that stop parsing when the parser inside fails. + +# Problem with current solution +We need to be dilligent in placing critical blocks. + +# Another approach +We could take a page from LR parsers and instead of driving the parsing from the parsers themselves we invert the flow, making the input drive the parsing. This results in a breadth first parsing, where the one-of block parses one step of every of it's children and decides which to keep. In the event there's only one children left, it is critical, so a failure stops parsing and reports the error. + +# How to select between BF and DF when parsing +Add an extra argument to parser type (lambda (input)) for the parsing mode. +one-of will be BF and comp will be DF. +This means that the immediate children of a one-of block will be parsed step by step. + +# Implementing BF in bind +Instead of calling the resulting function with the input, return it. +It will be the caller's responsibility to call the resulting lambda with the appropiate input. +This will give control to the one-of block to call the next parsing step. + |