QString: further optimize multiArg

This improves the previous algorithm in two aspects:

1. It parses the format string in a single pass instead of
   two, as was previously the case, storing all the
   information the result construction pass needs in a
   pair<QStringRef, int>.

This alone would slow the algorithm down, but without it, it's
hard to make the following work:

2. It avoids reallocations by calculating the size of the
   result and using the Qt::Uninitialized QString ctor.

This speeds up the following test case

    QString str("%1 %2 %3 %4 %5 %6 %7 %8 %9 foo %10 %11 bar"); // not timed
    // (arguments converted to QString outside loop)
    str = str.arg("one", "2", "3", "4", "5", "6", "7", "8", "9");
    str = str.arg("ahoy", "there");

from tst_qstring, which heavily favors the previous result
construction loop due to the tiny (but non-zero) amount of
verbatim text in between placeholders, by 25%:

  Qt 5.4:      1.50µs
  no map:      0.77µs (7b5ba56b0ab55fcaf79fbf9aad70bf767c938e15)
  this change: 0.58µs

Change-Id: I41ec86b7a21b4b25b3bc669ff2e3b2cc73513597
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>

0 files changed, 0 insertions, 0 deletions