wikimedia / parsoid / 2673

85%

push

travis-ci

Deal with JS/PHP differences wrt encoding/decoding of numerical keys

* JS always uses strings for keys, even for numerical keys.

* PHP always uses integers for numerical keys, even if started
  out as a string.

* See REPL output below.

[subbu@earth:~/work/wmf/parsoid/src/Html2Wt] node
> Object.keys(JSON.parse(JSON.stringify({1: 5, "2": 5})))
[ '1', '2' ]
> [subbu@earth:~/work/wmf/parsoid/src/Html2Wt] psysh
>>> array_keys(json_decode(json_encode([1 => 5, "2" => 5]), true));
=> [
     1,
     2,
   ]

* So, unlike JS code, remove the (string) cast from the comparison of
  keys in the PHP code since it will fail strict checks.

* Some other minor cleanup. We rely on data-parsoid info that Parsoid
  generates in a number of places in the codebase and don't always
  try to be conservative about bad data-parsoid input. If we want to be
  robust in the presence of corrupt data-parsoid objects, we need to do
  a code audit and/or throw a bunch of randomly generated input at
  Parsoid and fix code.

Change-Id: I876da8ce1

9528 of 12002 branches covered (79.39%)

14904 of 17619 relevant lines covered (84.59%)

90406.8 hits per line