notes

;; -*- mode: outline; outline-regexp: " *[-\+]"; indent-tabs-mode: nil; fill-column: 120 -*-

^c^d hide subtree
^c^s show subtree

Strict Dump Performance
JSON::Ext.dump 20000 times in 0.173 seconds or 115493.546 dump/sec.
Oj:strict.dump 20000 times in 0.050 seconds or 401475.753 dump/sec.

Summary:
   System  time (secs)  rate (ops/sec)
   ---------  -----------  --------------
   Oj:strict       0.050      401475.753
   JSON::Ext       0.173      115493.546

Comparison Matrix
(performance factor, 2.0 means row is twice as fast as column)
           Oj:strict  JSON::Ext
           ---------  ---------  ---------
           Oj:strict       1.00       3.48
           JSON::Ext       0.29       1.00

- big decimal
 - just in custom mode, maybe in strict?

 - encode_big_decimal_as_string
  - add as fix for rails change


- rails
   
 - Process::Status
 - ActiveSupport::TimeWithZone
 - ActiveModel::Errors
 - ActiveSupport::Multibyte::Chars
 - ActiveRecord::Relation

- debug
 - instrument for parsing floats/bigdecimal
  - branch

 - option to allow invalid unicode through
 - unit tests for 32 bit - test_float_parse
  - look at RUBY_PLATFORM maybe?
 - look at env var for OJ_DEBUG when doing ruby extconf.rb
 - add printf for keys spots in parse and dump
 
- invalid character option
 - don't check options
  - replace with best attempt such as \ud83d => \xED\xA0\xBD
  - that would be the normal replacement
 - allow_invalid_unicode

- streaming parser for scp and saj

---------------------------
Tried a separate thread for the parser and the results were poor. The parsing is
10% to 15% of the total so splitting ruby calls and c does not help much and the
overhead of swapping data was too high. It was not possible to split ruby calls
into both threads due to not getting a lock on the ruby environment for object
creation.