With this, the stack doesn't grow with every use of <|>, and alternatives are forgotten once we have committed to one branch. This should improve space usage and enable parsing of large documents without a stack overflow. I tested this new implementation with Bibdris. Performance seems to be on par with the old implementation.
ziman
commented
10 years ago
With this, the stack doesn't grow with every use of
<|>, and alternatives are forgotten once we have committed to one branch. This should improve space usage and enable parsing of large documents without a stack overflow. I tested this new implementation with Bibdris. Performance seems to be on par with the old implementation.