"""Efficiently compare files, boolean outcome only (equal / not equal).
Tricks (used in this order):
- Files with identical type, size & mtime are assumed to be clones
- Files with different type or size cannot be identical
- We keep a cache of outcomes of earlier comparisons
- We don't fork a process to run 'cmp' but read the files ourselves
def cmp(f1
, f2
, shallow
=1):
"""Compare two files, use the cache if possible.
Return 1 for identical files, 0 for different.
Raise exceptions if either file could not be statted, read, etc."""
s1
, s2
= sig(os
.stat(f1
)), sig(os
.stat(f2
))
if s1
[0] != 8 or s2
[0] != 8:
# Either is a not a plain file -- always report as different
# type, size & mtime match -- report same
if s1
[:2] != s2
[:2]: # Types or sizes differ, don't bother
# types or sizes differ -- report different
# same type and size -- look in the cache
cs1
, cs2
, outcome
= cache
[key
]
if s1
== cs1
and s2
== cs2
:
# cached signatures match
# stale cached signature(s)
cache
[key
] = s1
, s2
, outcome
"""Return signature (i.e., type, size, mtime) from raw stat data
0-5: st_mode, st_ino, st_dev, st_nlink, st_uid, st_gid
6-9: st_size, st_atime, st_mtime, st_ctime"""
"""Compare two files, really."""
bufsize
= 8*1024 # Could be tuned