forked from PyTables/PyTables
-
Notifications
You must be signed in to change notification settings - Fork 0
/
THANKS
86 lines (60 loc) · 3.17 KB
/
THANKS
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
March 2009
We would like to thank the people have contributed directly or
indirectly to PyTables.
Scott Prater for editing the user's manual in order to make it more
readable in english, as well as conducting the tests of PyTables on
Solaris.
Alan McIntyre for porting PyTables to Windows.
John Nielsen for suggesting improvements and delivering code for
completely avoid the recursion algorithms and allowing pytables to
bypass the ~1000 levels of deepness that Python recursion limit
imposed.
Tom Hedley for providing a nice patch for supporting complex datatypes
for Arrays, Errays and VLArrays. This was the root for the support of
complex types in Tables as well.
Shack Toms for providing a Python version of the nextafter and
nextafterf math functions that despite the fact they are standard in
C99 standard, they are not at the official places in Microsoft VC++
6.x nor VC++ 7.x.
Jeff Whitaker for providing the NetCDF module and the utility for
converting netCDF files to HDF5 (nctoh5).
Norbert Nemec for providing several interesting patches.
Andrew Straw for suggesting to bracket the most intensive parts of
PyTables with BEGIN_ALLOW_THREADS and END_ALLOW_THREADS. That will
allow much better performance of PyTables apps in mutiprocessors
platforms.
Antonio Valentino for providing several patches for supporting
native multidimensional attributes and the CArray object.
Ashley Walsh, for reporting several problems and fixes. It has helped
testing OSX platform, specially UCL compressor issues.
Russel Howe, for reporting and providing an initial patch for a nasty
memory leak when reading VLArray types.
The HDF5 team at NCSA for making such an excellent library for data
persistence, and specially Pedro Vicente, Quincey Koziol and Elena
Pourmal, for quickly including my suggested patches to the HDF5_HL and
solving the reported bugs in HDF5 library.
Todd Miller and Perry Greenfield for promptly helping me to understand
many of the intricacies of the numarray package and Jin-chung Hsu for
discussions on recarray module (now numarray.records module). They
have been very receptive and promptly worked-out most of the
improvements in numarray (specially in the records module) that were
necessary for PyTables.
Travis Oliphant for its impressive work and responsiveness with NumPy.
Evan Prodromou for his lrucache package, a very sleek implementation
of an LRU queue. He had a very helpful attitude with the licensing
and technical issues.
Gerard Vermeulen for Windows/MSVS-2005 testing.
Enric Cervera for testing the binaries for MacOSX/Intel.
Daniel Bungert, Steve Langasek and Alexandre Fayolle for their support
in creating Debian packages for PyTables.
Greg Ewing for writing the excelent Pyrex tool and allowing to beginners
like me to quickly and safely start writing Python extensions. He was
also very responsive about questions on Pyrex.
Stefan Behnel, Robert Bradshaw, and Dag Sverre Seljebotn for their
impressive work with Cython.
Andrew Collette, for his excellent work on the h5py project, from
which PyTables starts to stole ideas (and code too ;-).
Guido, you know who.
And last, but definitely not least!,
To those companies that are supporting the PyTables project with
contracts.