If settings.ini is present next to the executable, will run in portable mode.
This results in settings, data, and cache all being in same folder as dupeGuru.
Computing 3 hash samples for files less than 3MiB (3 * CHUNK_SIZE) is not efficient since spans of later samples would overlap a previous one.
Therefore we can simply return the hash of the entire small file instead.
- Upgrade requirements to specify more current minimums
- Remove compatability code from sphinxgen for old versions
- Upgrade pyinstaller to a minimum version that works with latest macOS
- Pull updates from Transifex
- Add Turkish
- Sort language lists in code
- Remove old locale conversion code as it appears to work correctly on
windows without different conversions.
* Some refs appeared in the dupes-only view after a re-prioritization was done a second time.
* It seems the core.Results.__dupes list was not properly updated whenever core.app.Dupeguru.reprioritize_groups() -> core.Results.sort_dupes() was called.
When a re-prioritization is done, some refs became dupe, and some dupes became ref in their place. So we need to update the new state of the internal list of dupes kept by the Results object, instead of relying on the outdated cached one.
* Fix#757.
* Instead of keeping md5 samples separate, merge them as one hash computed from the various selected chunks we picked.
* We don't need to keep a boolean to see whether or not the user chose to optimize; we can simply compare the value of the threshold, since 0 means no optimization currently active.
* Big files above the user selected threshold can be partially hashed in 3 places.
* If the user is willing to take the risk, we consider files with identical md5samples as being identical.
* Doing a full match should be safer to avoid partial results which would result in overly aggressive filtering.
* Add new tests to test suite to cover this issue.
* Fixes#903.
* Accents are getting removed from Unicode characters to generate similar "words".
* Non-latin characters which cannot be processed that way (eg. japanese, greek, russian, etc.) should not be filtered out at all otherwise files are erroneously skipped or detected as dupes if only some characters make it passed the filter.
* Starting from an arbitrary unicode codepoint (converted to decimal), above which we know it is pointless to try any sort of processing, we leave the characters as is.
* Fix#878.
* Fixes "'DetailsPanel' object has no attribute '_table'" error on shutdown if the Results table is updated (item removed) while the Details Dialog is shown as a floating window.
* It seems that QApplication.quit() triggers some sort of refresh on the floating QDockWidget, which in turn makes calls to the underlying model that is possibly being destroyed, ie. there might be a race condition here.
* Closing or hiding the QDockWidget before the cal to quit() is a workaround. Similarly, this is already done in the quitTriggered() method anyway.
* This fixes#857.
