robertvokac/bitrot
Archived
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Activity

v1.0.0

Browse Source
This commit is contained in:
Łukasz Langa 2020-05-18 00:15:24 +02:00
parent 0dc3390b7f
commit 67e7b8c904
No known key found for this signature in database
GPG Key ID: B26995E310250568
2 changed files with 27 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

46
README.rst
View File

@ -22,27 +22,28 @@ will report all errors, e.g. files that changed on the hard drive but
still have the same modification date. still have the same modification date.
All paths stored in ``.bitrot.db`` are relative so it's safe to rescan All paths stored in ``.bitrot.db`` are relative so it's safe to rescan
a folder after moving it to another drive. a folder after moving it to another drive. Just remember to move it in
a way that doesn't touch modification dates. Otherwise the checksum
database is useless.
Performance Performance
----------- -----------
Obviously depends on how fast the underlying drive is. Since bandwidth Obviously depends on how fast the underlying drive is. Historically
for checksum calculations is greater than your drive's data transfer the script was single-threaded because back in 2013 checksum
rate, even when comparing mobile CPUs vs. SSD drives, the script is calculations on a single core still outran typical drives, including
single-threaded. the mobile SSDs of the day. In 2020 this is no longer the case so the
script now uses a process pool to calculate SHA1 hashes and perform
`stat()` calls.
No rigorous performance tests have been done. Scanning a ~1000 files No rigorous performance tests have been done. Scanning a ~1000 file
totalling ~4 GB takes 20 seconds on a 2015 Macbook Air (SM0256G SSD). directory totalling ~5 GB takes 2.2s on a 2018 MacBook Pro 15" with
This is with cold disk cache. a AP0512M SSD. Back in 2013, that same feat on a 2015 MacBook Air with
a SM0256G SSD took over 20 seconds.
Some other tests back from 2013: a typical 5400 RPM laptop hard drive On that same 2018 MacBook Pro 15", scanning a 60+ GB music library takes
scanning a 60+ GB music library took around 15 minutes. On an OCZ 24 seconds. Back in 2013, with a typical 5400 RPM laptop hard drive
Vertex 3 SSD drive ``bitrot`` was able to scan a 100 GB Aperture library it took around 15 minutes. How times have changed!
in under 10 minutes. Both tests on HFS+.
If you'd like to contribute some more rigorous benchmarks or any
performance improvements, I'm accepting pull requests! :)
Tests Tests
----- -----
@ -54,17 +55,22 @@ file in the `tests` directory to run it.
Change Log Change Log
---------- ----------
0.9.3 1.0.0
~~~~~ ~~~~~
* significantly sped up execution on solid state drives by using
a process pool executor to calculate SHA1 hashes and perform `stat()`
calls; use `-w1` if your runs on slow magnetic drives were
negatively affected by this change
* sped up execution by pre-loading all SQLite-stored hashes to memory
and doing comparisons using Python sets
* all UTF-8 filenames are now normalized to NFKD in the database to * all UTF-8 filenames are now normalized to NFKD in the database to
enable cross-operating system checks enable cross-operating system checks
* the SQLite database is now vacuumed to minimize its size * the SQLite database is now vacuumed to minimize its size
* sped up execution by pre-loading all SQLite-stored hashes to memory
and doing comparisons using Python sets
* bugfix: additional Python 3 fixes when Unicode names were encountered * bugfix: additional Python 3 fixes when Unicode names were encountered
0.9.2 0.9.2
@ -201,4 +207,4 @@ improvements by
`Reid Williams <rwilliams@ideo.com>`_, `Reid Williams <rwilliams@ideo.com>`_,
`Stan Senotrusov <senotrusov@gmail.com>`_, `Stan Senotrusov <senotrusov@gmail.com>`_,
`Yang Zhang <mailto:yaaang@gmail.com>`_, and `Yang Zhang <mailto:yaaang@gmail.com>`_, and
`Zhuoyun Wei <wzyboy@wzyboy.org>`_. `Zhuoyun Wei <wzyboy@wzyboy.org>`_.

2
src/bitrot.py
View File

@ -45,7 +45,7 @@ from concurrent.futures import ProcessPoolExecutor, wait, as_completed
DEFAULT_CHUNK_SIZE = 16384 # block size in HFS+; 4X the block size in ext4 DEFAULT_CHUNK_SIZE = 16384 # block size in HFS+; 4X the block size in ext4
DOT_THRESHOLD = 200 DOT_THRESHOLD = 200
VERSION = (0, 9, 2) VERSION = (1, 0, 0)
IGNORED_FILE_SYSTEM_ERRORS = {errno.ENOENT, errno.EACCES} IGNORED_FILE_SYSTEM_ERRORS = {errno.ENOENT, errno.EACCES}
FSENCODING = sys.getfilesystemencoding() FSENCODING = sys.getfilesystemencoding()