When requesting a list of text/uri-list, use the much faster QMimeData
urls() method.
The unittests pass (the desktop: protocol) and
CCBUG: 342056
is half solved. The paste speed is as fast as drag&drop in local files
and with fish: files.
The lock in X11 plasma or kwin still needs another patch.
Details
Details
- Reviewers
dfaure - Group Reviewers
Frameworks - Commits
- R244:18e4d245d3d5: Use the much faster urls() method from QMimeData
Select 2000 files in one folder, cut and paste them in another disk.
Diff Detail
Diff Detail
- Repository
- R244 KCoreAddons
- Lint
Lint Skipped - Unit
Unit Tests Skipped
Comment Actions
Ah I guess it's faster because QMimeData skips the encoding/decoding via QByteArray when both drag and drop are in the same process? That's not something we can do ourselves (for the "kde uri list" mimetype), unless we call the protected QMimeData::retrieveData, I guess. Hmm.
| src/lib/io/kurlmimedata.cpp | ||
|---|---|---|
| 76 | But then why call this at all, if you don't need the bytearray? if (mimeData->hasUrls()) | |
Comment Actions
- Use the much faster urls() method from QMimeData
Using hasUrls() and fixed a typo in the comment
| src/lib/io/kurlmimedata.cpp | ||
|---|---|---|
| 74 | if firstMimeType == "text/uri-list" (see above, PreferLocalUrls), it's still going to be slow, no? Should this be handled generally here by introducing a helper that checks the mime type and then uses either urls() directly or call data as before? also, what happens if PreferLocalUrls is set, but no text/uri-list data is available, could the old way of getting the data for the s_kdeUriListMime have worked more reliably? | |
Comment Actions
Something like this?
if firstMimeType == "text/uri-list" (see above, PreferLocalUrls), it's still going to be slow, no? Should this be handled >generally here by introducing a helper that checks the mime type and then uses either urls() directly or call data as >before?
also, what happens if PreferLocalUrls is set, but no text/uri-list data is available, could the old way of getting the data >for the s_kdeUriListMime have worked more reliably?
If PreferLocalUrls then firstMimeType = text/url-list
if firstMimeType == text/url-list apply urls() method
The test hash both desktop: and file: urls in the same MimeType and it passes.
| src/lib/io/kurlmimedata.cpp | ||
|---|---|---|
| 75 | This can't be right, it would mean ba is ignored when firstMimeType==text/uri-list (which would mean the call to data() above was for nothing). Since the method to call (data() or urls()) depends on the mimetype, either the whole idea of swapping the mimetypes has to be dropped, or as Milian suggests, a helper with an if() should encapsulate this (so we can call it in both places) (but that's more string comparisons....). To make it fast I'd do, well, OK the code wouldn't fit into this margin, let's make a separate RR: https://phabricator.kde.org/D10257 | |