smmap is not marked as Python 3 compatible on PyPi but actually is Python 3 compatible.

[schryer]

I installed smmap in Python 3.4 and ran the tests. They all passed (see below).

Is there another reason why this package does not come up green on this webpage:
http://py3readiness.org/

The authors of GitPython and gitdb claim that they cannot provide Python 3 support
unless smmap is Python 3 compatible, but it already seems to be and thus should be
marked as providing Python 3 support on PyPi.

optimal: Made 100 random single byte accesses to buffer created from <class 'str'> reading a total of 0.000100 mb in 0.282117 s (0.000354 mb/s)
optimal: Made 100 random slice accesses to buffer created from <class 'str'> reading a total of 205.965117 mb in 2.003208 s (102.817655 mb/s)
worst case: Made 100 random single byte accesses to buffer created from <class 'str'> reading a total of 0.000100 mb in 0.009837 s (0.010165 mb/s)
worst case: Made 100 random slice accesses to buffer created from <class 'str'> reading a total of 193.733239 mb in 2.974128 s (65.139504 mb/s)
static optimal: Made 100 random single byte accesses to buffer created from <class 'str'> reading a total of 0.000100 mb in 0.286831 s (0.000349 mb/s)
static optimal: Made 100 random slice accesses to buffer created from <class 'str'> reading a total of 190.569213 mb in 1.856029 s (102.675772 mb/s)
optimal: Made 100 random single byte accesses to buffer created from <class 'int'> reading a total of 0.000100 mb in 0.295136 s (0.000339 mb/s)
optimal: Made 100 random slice accesses to buffer created from <class 'int'> reading a total of 183.239070 mb in 1.812193 s (101.114523 mb/s)
worst case: Made 100 random single byte accesses to buffer created from <class 'int'> reading a total of 0.000100 mb in 0.008187 s (0.012214 mb/s)
worst case: Made 100 random slice accesses to buffer created from <class 'int'> reading a total of 201.191009 mb in 3.076563 s (65.394738 mb/s)
static optimal: Made 100 random single byte accesses to buffer created from <class 'int'> reading a total of 0.000100 mb in 0.295137 s (0.000339 mb/s)
static optimal: Made 100 random slice accesses to buffer created from <class 'int'> reading a total of 191.903437 mb in 1.871780 s (102.524573 mb/s)
..<class 'smmap.mman.StaticWindowMapManager'>: Read 110 mb of memory with 5000 random on cursor initialized with <class 'str'> accesses in 1.002509s (109.761307 mb/s)
<class 'smmap.mman.StaticWindowMapManager'>: Read 67 mb of memory with 5000 random on cursor initialized with <class 'int'> accesses in 0.640364s (105.108681 mb/s)
<class 'smmap.mman.SlidingWindowMapManager'>: Read 190 mb of memory with 5000 random on cursor initialized with <class 'str'> accesses in 2.169544s (87.846437 mb/s)
<class 'smmap.mman.SlidingWindowMapManager'>: Read 190 mb of memory with 5000 random on cursor initialized with <class 'int'> accesses in 2.110421s (90.109472 mb/s)

.......

Ran 9 tests in 21.284s

OK

[Byron]

Confirmed, even though there is a performance regression - python 3 is 7 times slower than python 2.
I am now working on python 3 compatibility and respective fixes, will keep you posted here.

[Byron]

A new release was made on pypi, reflecting the py3 compatibility.
The performance regression was fixed, even though py3 is still about 1.5 times slower in handling raw byte data.