Accelerator readme doesn't give useful examples of acceleration

[rcrogers]

https://github.com/kschiess/parslet/blob/9afeb3ca5ce7f83786f9a78cb5dca06a199b29f3/qed/accelerators.md#bindings-and-actions

Not sure if this is an oversight or just unfinished, but I couldn't find an example of how an accelerator can actually speed up parsing, e.g. an example implementation of GobbleUp. Will it always involve creating a custom Atom class?

These would also be useful:

• how to find hot spots
• approximate performance numbers (for an example scenario)
• an example where an accelerator's rule-matching capability has an advantage over just referencing custom atoms in the parser (GobbleUp appears to be a 1-to-1 substitution that doesn't make the parser substantially "dirtier")

[rcrogers]

I found GobbleUp: https://github.com/kschiess/parslet/blob/66e242841ea27554a64f42d52093c1f41bc56e1e/qed/applique/gobbleup.rb

Do accelerators just boil down to finding a faster method like StringScanner#check_until?

[kschiess]

Honestly, I've ran into a design roadblock with accelerators - and have let it rest since then. The issue is that we want to support lazy evaluation across rule barriers - and that collides with acceleration.

The most potential for speeding up parsers I see in creating custom 'partial' parsers aka atom classes. But if you really need to go fast - don't use ruby. Simple as that.

[rcrogers]

All good to know, thanks for the response!

[kschiess]

Welcome! (closing this)