[Bio] / Numeric / changes.txt Repository:
ViewVC logotype

Annotation of /Numeric/changes.txt

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1.1.1.1 - (view) (download)

1 : efrank 1.1 Version 23.7 January 2005
2 :     [1075953] added franf to fix problems on AMD64.
3 :     [1027480] this should have been fixed before.
4 :     [scipy] fixed reshape to convert dimensions to PyArray_LONG to avoid typecasting problem.
5 :     [eric jones] added empty to generate unitialized array
6 :     [user ] added support for vecLib on MacOS X in setup.py
7 :     [unreported] added UfuncType to Numeric to Parallel ArrayType
8 :    
9 :     Version 23.6
10 :     [unreported] fixed argmin to work with Unsigned Array Types
11 :     [unreported] fixed UINT_to_XXX methods so input is assumed unsigned int
12 :     [unreported] fixed UINT conversion rules and inappropriate casting to unsigned integers
13 :     from signed integers
14 :     [build ] fixed tarball so same tarball for multiple versions.
15 :    
16 :    
17 :     Version 23.5 Sept 2004
18 :     [unreported] changed arrayfnsmodule to look for PyArray_LONG on input not PyArray_INT
19 :     [perez ] fixed so that matrixmultiply is always dot (even if BLAS optimized)
20 :    
21 :     Version 23.4 Sept 2004
22 :     [unreported] fixed setup.py file so ATLAS not default
23 :     [unreported] altered coercion so that LONG is not cast to INT safely
24 :    
25 :     Version 23.3 June 2004
26 :     [unreported] misspelled UFunc_Type in ufunc API
27 :     [ 843510 ] ubyte abs giving incorrect results
28 :     [ 919540 ] added weakreferences to Numeric Object
29 :     [unreported] fixed segfault due to unchecked NULL in get_type
30 :     [unreported] added faster array(a) when a is already an array
31 :     [unreported] a can now be interpreted as a scalar as long as it only
32 :     has one element (instead of just zero-dimensions).
33 :     changed compress keyword from dimension to axis to match
34 :     everyother function in the library.
35 :    
36 :     Version 23.1 August 2003
37 :     [unreported] fixed assignment to matrix.
38 :     [unreported] added check so that non-character array cannot be interpreted
39 :     as a character buffer.
40 :     [unreported] array([m,m,m]) when m is a rank-0 array now works as expected.
41 :     Patches:
42 :     [ 782001 ] Add true and floor divide to MA
43 :     [ 781215 ] true_divide bug
44 :     [ 776991 ] multiplying small UINT values gives random results
45 :     [ 776467 ] remove string exceptions
46 :    
47 :     Version 23 March 2003
48 :    
49 :     Important notice:
50 :    
51 :     Two packages have been removed from optional ones: PropertiedClasses, kinds.
52 :     MA has been rewritten to use standard property and will not work for
53 :     ancient Pythons. (Pre 2.1, I think). Use the MA / Propertied Classes
54 :     from Numeric 22 if you can't use this one.
55 :    
56 :     The kinds package (subject of PEP-0242) will be released as a separate
57 :     package shortly. PEP-0242 was withdrawn because this facility did not seem
58 :     to be worth putting in the standard library, but kinds is correct as is.
59 :    
60 :     [ 695200 ] Richard Everson (R.M.Everson@exeter.ac.uk) has donated a dotblas
61 :     Package that gets Numeric to use optimized BLAS libraries for dot
62 :     innerproduct, and vdot --- a conjugate vector dot product he introduced.
63 :     setup.py must be altered by the user in a manner similar to the alterations
64 :     to use optimized BLAS for LinearAlgebra
65 :    
66 :     [675777] new-style classes as objects in a sequence were not being detected
67 :     correctly by array_objecttype. Corrected array_objecttype to handle
68 :     them. (Oliphant)
69 :    
70 :     [contribution] Fernando Perez has donated a revised version of the tutorial
71 :     file view.py that seems to be less likely to hang the interpreter.
72 :    
73 :     [68392923] dimensions+scalar -> crash (jneb)
74 :     Found divergent value of MAX_DIMS in ufuncobject.c;
75 :     The value was 20 there and 40 in two other places. But 40 is ludicrous,
76 :     we would never have that much memory available. Changed
77 :     them all to 30.
78 :    
79 :     [ unreported ] Changed PY_VERSION_HEX check for version 2.2
80 :     to 0x0202000 as it should have been
81 :     so that true_division numeric ops can be supported
82 :    
83 :     [ 614808 ] Inconsistent use of tabs and spaces
84 :     Fixed as suggested by Jimmy Retzlaff
85 :     LinearAlgebra.py
86 :     Matrix.py
87 :     RNG/__init__.py
88 :     RNG/Statistics.py
89 :    
90 :     [ 621032 ] needless work in multiarraymodule.c
91 :     Fixes suggested by Greg Smith applied.
92 :     Also recoded OBJECT_DotProduct to eliminate a warning error.
93 :    
94 :     [ 630584 ] generalized_inverse of complex array
95 :     Fix suggested by Greg Smith applied.
96 :    
97 :     [ 652061 ] PyArray_As2D doesn't check pointer.
98 :     Fix suggested by Andrea Riciputi applied.
99 :    
100 :     [ 655512 ] inverse_real_fft incorrect many sizes
101 :     Fix given by mbriest applied.
102 :    
103 :     [unreported] a.real increased reference count of a and raised error when
104 :     a is not complex.
105 :     Fixed to apparent intended behavior of returning an array
106 :     with the same data. (Oliphant)
107 :    
108 :     Patch for 64-bit machines applied. Appears to work ok on 32 bit but don't
109 :     have the machine to test the patch.(Dubois)
110 :    
111 :     [ 627771 ] matrixmultiply broken for non-contig (fixed, test case added)
112 :     (Greg Smith)
113 :    
114 :     [unreported] Fixed ArrayPrinter when NaN's show up in Float32
115 :     precision.
116 :    
117 :     [ 545336 ] Bug in RandomArray.randint
118 :     Changed the function ranf to type double (Chuck Harris)
119 :     Harris is probably right that all floats should be double in this module
120 :     but it may be this has performance or storage consequences that would
121 :     bite somebody. I support RNG, not this one. -- Dubois
122 :    
123 :     Version 22.0 August , 2002
124 :     a. Changed multiarraymodule functions to accept keywords where documentation implies
125 :     it through the use of optional variables.
126 :     Specifically in multiarray: zeros, take, transpose, repeat,
127 :     set_string_function, cross_correlate.
128 :     in ufuncobject: reduce and accumulate now take keyword arguments
129 :     for the optional axis argument.
130 :     b. Added support for unsigned shorts 'w' and unsigned ints 'u'
131 :     -- Travis Oliphant with help from Darren Hart and F. Oliver Gathmann.
132 :     Increased max permissible iterations in SVD for supplied lapack. -- Dubois
133 :     Recoded RandomArray.randint to try to see if we can work around bug
134 :     on some platforms. -- Dubois
135 :    
136 :     Version 21.3 June 8, 2002
137 :     Fixed bugs:
138 :     [ #557927 ] fixed matrix slice assignment
139 :     [ #552922 ] added check for correct datatype in .astype() method.
140 :     Created new API PyArray_ValidType to handle this check
141 :     here as well as in multiarraymodule.c
142 :     [ #551808 ] fixed segfault with unicode array (Travis O.)
143 :     [ #559511 ] MLab.std now works for axis != 0 (Travis O.)
144 :     [ #542979 ] sum returns exception tuple
145 :    
146 :     [ #528328 ] true division operators used to return single precision on division of
147 :     integers and longs --- now defaults to double precision (but only on int and
148 :     long division --- still single-precision for ubyte, short, and byte division.
149 :     [ none ] arange(start, end, step) slightly different near end points than
150 :     start + arange(0, N)*step where N is the length.
151 :     [ none ] a = zeros(2,'D'); a[0] = array(0.0+0.6j) would not work. (Assigning a
152 :     rank-0 array did not work for CFLOAT_setitem or CDOUBLE_setitem.
153 :     [ 530688 ] Python crash when transposing array (Walter Moreira)
154 :    
155 :     Version 21.0 March 13, 2002
156 :     Fixed bugs:
157 :     [ #482603 ] Memory leak in MA/Numeric/Python
158 :     Reported by Reggie Dugard. Turned out to be
159 :     *two* memory leaks in one case in a routine in Numeric,
160 :     array_objectype. (Dubois)
161 :     [ none ] if vals was a null-array array([]) putmask and put would
162 :     crash. Fixed with check.
163 :     [ #469951 ] n = n1[0] gives array which shares dimension of n1 array.
164 :     This causes bugs if shape of n1 is changed (n didn't used
165 :     to have it's own dimensions array (Travis Oliphant)
166 :     [ #514588 ] MLab.cov(x,x) != MLab.cov(x) (Travis Oliphant)
167 :     [ #518702 ] segfault when invalid typecode for asarray (Travis Oliphant)
168 :     [ #497530 ] MA __getitem__ prevents 0 len arrays (Reggie Duggard)
169 :     [ #508363 ] outerproduct of noncontiguous arrays (Martin Wiechert)
170 :     [ #513010 ] memory leak in comparisons (Byran Nollett)
171 :     [ #512223 ] Character typecode not defined (Jochen Kupper)
172 :     [ #500784 ] MLab.py diff error (anonymous, fixed by Dubois)
173 :     [ #503741 ] accuracy of MLab.std(x) (Katsunori Waragai)
174 :     [ #507568 ] overlapping copy a[2:5] = a[3:6]
175 :     Change uses of memcpy to memmove which allows overlaps.
176 :     [ numpy-Patches-499722 ] size of buffer created from array is bad (Michel Sanner).
177 :     [ #502186 ] a BUG in RandomArray.normal (introduced by last bug fix in 20.3)
178 :     (Katsunori Waragai).
179 :    
180 :     Fixed errors for Mac (Jack Jensen).
181 :    
182 :     Make rpm's properly, better Windows installers. (Gerard Vermeulen)
183 :     Added files setup.cfg; setup calculates rpm_install.sh to use current Python.
184 :     New setup.py, eliminate setup_all.py. Use os.path.join everywhere. Revision in b6
185 :     added file README.RPM, further improvements.
186 :    
187 :     Implement true division operations for Python 2.2. (Bruce Sherwood)
188 :     Note: true division of all integer types results in an array of floats,
189 :     not doubles. This decision is arbitrary and there are arguments either way,
190 :     so users of this new feature should be aware that the decision may
191 :     change in the future.
192 :    
193 :     New functions in Numeric; they work on any sequence a that can be converted to a
194 :     Numeric array. Similar change to average in MA. (Dubois)
195 :    
196 :     def rank (a):
197 :     "Get the rank of a (the number of dimensions, not a matrix rank)"
198 :    
199 :     def shape (a):
200 :     "Get the shape of a"
201 :    
202 :     def size (a, axis=None):
203 :     "Get the number of elements in a, or along a certain axis."
204 :    
205 :     def average (a, axis=0, weights=None, returned = 0):
206 :     """average(a, axis=0, weights=None)
207 :     Computes average along indicated axis.
208 :     If axis is None, average over the entire array.
209 :     Inputs can be integer or floating types; result is type Float.
210 :    
211 :     If weights are given, result is:
212 :     sum(a*weights)/(sum(weights))
213 :     weights must have a's shape or be the 1-d with length the size
214 :     of a in the given axis. Integer weights are converted to Float.
215 :    
216 :     Not supplying weights is equivalent to supply weights that are
217 :     all 1.
218 :    
219 :     If returned, return a tuple: the result and the sum of the weights
220 :     or count of values. The shape of these two results will be the same.
221 :    
222 :     raises ZeroDivisionError if appropriate when result is scalar.
223 :     (The version in MA does not -- it returns masked values).
224 :     """
225 :    
226 :    

MCS Webmaster
ViewVC Help
Powered by ViewVC 1.0.3