Mailman 3 September 2009 - NumPy-Discussion

Compound conditional indexing
by Gökhan Sever 30 Sep '09

30 Sep '09

Hello, How to conditionally index an array as shown below : a = arange(10) a[5<a<8] to get array([6,7]) I can't do this with where either. What is the cure for this? Thanks. -- Gökhan

4 5

numpy macosx10.5 binaries: compatible with 10.4?
by Russell E. Owen 30 Sep '09

30 Sep '09

All the official numpy 1.3.0 Mac binaries are labelled "macosx10.5". Does anyone know if these are backwards compatible with MacOS X 10.4 or 10.3.9? -- Russell

3 2

how does numpy handle views and garbage collection?
by Chris Colbert 30 Sep '09

30 Sep '09

Lets say I have function that applies a homogeneous transformation matrix to an Nx3 array of points using np.dot. since the matrix is 4x4 I have to add a 4 column of ones to the array so the function looks something like this: def foo(): <--snip--> pts = np.column_stack((Xquad, Yquad, Zquad, np.ones(Zquad.shape))) transpts = np.dot(transmat, pts.T).T return transpts[:,:3] Since i'm returning just the view of the array, I imagine python doesnt garbage collect transpts once the function returns and falls out of scope (because numpy has increfed it in the view operation?). So in essence, I still have that whole column of ones hanging around wasting memory, is that about right? Cheers, Chris

2 1

MemoryError with fancy indexing
by Sebastian Haase 30 Sep '09

30 Sep '09

Hi, Maybe someone could explain to me what is going on here!? >>> N.who() Name Shape Bytes Type ============================================================ SLmap 2048 x 2048 16777216 int32 SLmap_fast 2048 x 2048 16777216 int32 SLmap_slow 2048 x 2048 16777216 int32 <snip> >>> SLmap[:]=0 >>> SLmap[SLmap_fast]|=2 Traceback (most recent call last): File "<input>", line 1, in <module> MemoryError >>> SLmap[SLmap_fast]=2 >>> SLmap[SLmap_slow]+=1 Traceback (most recent call last): File "<input>", line 1, in <module> MemoryError >>> SLmap[SLmap_fast.astype(N.bool)]+=2 >>> SLmap[:]=0 >>> SLmap[SLmap_fast.astype(bool)]|=2 >>> SLmap[SLmap_slow.astype(bool)]|=1 >>> Why do I run into memory problems using an int32 array as index - while not having any problem when converting it first to dtype=bool ? (I'm on 64-bit Linux, having many GBs physical ram) Thanks, Sebastian Haase

1 0

max value of np scalars
by Neal Becker 30 Sep '09

30 Sep '09

I need the max value of an np scalar type. I had used this code: def get_max(is_signed, base_type, total_bits): print 'get_max:', is_signed, base_type, total_bits if is_signed: return (~(base_type(-1) << (total_bits-1))) else: print type(base_type (-1) << total_bits) return (~(base_type (-1) << total_bits)) This doesn't work for e.g., np.uint64. As the 'print' shows, get_max: False <type 'numpy.uint64'> 10 <type 'long'> The type of np.uint64 (-1) << 10 is not np.uint64, but long. This seems very strange to me. So, 2 questions. 1) Is this expected behavior? 2) How can I correctly implement get_max?

4 3

fixed_pt prototype using aggregation
by Neal Becker 30 Sep '09

30 Sep '09

I have a prototype for fixed_pt without using inheritance. I think I like it. Any thoughts?

2 2

__array_wrap__
by Neal Becker 30 Sep '09

30 Sep '09

fixed_pt arrays need to apply the overflow_policy after operations (overflow_policy could be clip, or throw exception). I thought __array_wrap__ would work for this, but it seems to not be called when I need it. For example: In [13]: obj Out[13]: fixed_pt_array([ 0, 32, 64, 96, 128]) In [14]: obj*100 < this should overflow enter: [ 0 32 64 96 128] << on entry into __array_wrap enter: [0 32 64 96 128] exit: [ 0 32 64 96 128] Out[14]: fixed_pt_array([ 0, 3200, 6400, 9600, 12800]) Apparantly, obj*100 is never passed to array_wrap. Is there another way I can do this?

5 12

making the distinction between -0.0 and 0.0..
by Christopher Barker 29 Sep '09

29 Sep '09

Hi folks, This isn't really a numpy question, and I'm doing this with regular old python, but I figure you are the folks that would know this: How do I get python to make a distinction between -0.0 and 0.0? IN this case, I'm starting with user input, so: In [3]: float("-0.0") Out[3]: -0.0 so python seems to preserve the "-". But: In [12]: float("-0.0") == float("0.0") Out[12]: True In [13]: float("-0.0") < float("0.0") Out[13]: False In [14]: float("0.0") > float("-0.0") Out[14]: False It doesn't seem to make the distinction between -0.0 and 0.0 in any of the comparisons. How can I identify -0.0? NOTE: numpy behaves the same way, which I think it should, but still... My back-up plan is to process the string first, looking for the minus sign, but that will require more changes than I'd like to the rest of my code... thanks, -Chris -- Christopher Barker, Ph.D. Oceanographer Emergency Response Division NOAA/NOS/OR&R (206) 526-6959 voice 7600 Sand Point Way NE (206) 526-6329 fax Seattle, WA 98115 (206) 526-6317 main reception Chris.Barker(a)noaa.gov

10 19

chebyshev module
by Charles R Harris 29 Sep '09

29 Sep '09

Hi all, I'm at the polishing stage on this module and at this point would like some input on the names. Yeah, a bit late ;) As it stands the module emulates the polynomial module in most things with the substitution of cheb for poly and the poly1d equivalent is Cheb1d. There are also a few deviations, instead of cheb there is chebfromroots, roots is chebroots, there are converters to,from polynomials, the ctor for Cheb1d only takes coefficients (not roots or other instances), and Cheb1d doesn't have the __array__ method. The supported coefficients arrays all have to be 1d, but the argument passed to chebval can be any array. Any suggestions are welcome at this time. I'll post the code soonish for further comment. Oh, and is it advisable to have a __copy__ (or copy) method? Any other functions? I don't want to do a complete workup, that would seem to belong to scipy. Chuck

2 1

where does numpy get its pow function?
by Chris Colbert 29 Sep '09

29 Sep '09

Does numpy use pow from math.h or something else? I seem to be having a problem with slow pow under gcc when building an extension, but it's not affecting numpy. So if numpy uses that, then there is something else i'm missing. Cheers! Chris

5 7