so much for lightweight markup
Rendering the same 6-paragraph, 65-line source markup file (reStructuredText or pandoc's Markdown flavor) into a 154-line HTML 5 target file:
py36-rst2html5: 1.05 seconds
py36-docutils: 0.65 seconds
hs-pandoc: 2.50 seconds
Rendering the same source file content through Jamal, "Just Another Macro Language," a perl script from 1997 & last updated 2010:
jamal: 0.05 seconds
These look tiny now, but what happens when I go over 100 pages then tweak the template?
so much for lightweight markup
py36-rst2html5 on a year-old Celeron Braswell CPU: 1.05 seconds.
py36-rst2html5 on a 10-year-old Core 2 Quad Kentsfield CPU: 0.55 seconds.
Same OS, same version, same input file, just different hardware. ... Wat.
@arielmt i'd kinda expect hs-pandoc at least to be much faster than jamal when you go over 100 pages? once you get past startup overhead, haskell should be quicker
@00dani I thought of that, but I was thinking it was a general caching issue, not invocation overhead. I had timed consecutive runs of pandoc, & the first run was over 3X slower than subsequent runs (7.96 s vs consistent 2.44-2.51 s just now in a repeat test). The fastest number is what I used.
so much for lightweight markup
Also would address some of your prior complaints like the title thing.
so much for lightweight markup
@arielmt Funky! Maybe something to do with memory access? Or even branch prediction? That could be quite fun to get to the bottom of..
so much for lightweight markup
This is for a static website, BTW.
I'm done with fast-changing dynamic frameworks & CMS all-in-ones with even half the dependencies of "Hello World" in Ruby on Rails.