diff options
author | Justin Luth <justin_luth@sil.org> | 2017-02-08 19:08:07 +0300 |
---|---|---|
committer | Bartosz Kosiorek <gang65@poczta.onet.pl> | 2017-02-10 17:13:53 +0000 |
commit | 09e9274fc080b471393b806617eb03124db67590 (patch) | |
tree | 8d98aae4ea6885502b9ada04e1f8e7fe2b378abf /m4 | |
parent | 07bcdbaed1427549e29af00107b99878b9a66a72 (diff) |
Optimize Excel GetOrCreateRow: reduce loops
for ( size_t nFrom = maRowMap.size(); nFrom <= nXclRow; ++nFrom )
This previous code worked best under the assumption
that every row is added to the map. However, the size of the
map actually has no correlation to the row numbers contained in it
when many rows are identical to each other (think silly formatting
and empty rows - related to tdf#105840)
Thus row 1,000,000 could occupy slot 2, and every access of that row
would then trigger nearly 1 million redundant loops.
Optimize:
-check to see if the row already exists - if so do nothing.
-existance of higher rows indicates there are no missing rows.
-build missing rows from the previously-mapped row.
Change-Id: Ib02520a1bf0f77b5ca0ec5ad3165ff7ea879515f
Reviewed-on: https://gerrit.libreoffice.org/34038
Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Bartosz Kosiorek <gang65@poczta.onet.pl>
Diffstat (limited to 'm4')
0 files changed, 0 insertions, 0 deletions