Re: [Numpy-discussion] Change default order to Fortran order
Hi, On Mon, Aug 3, 2015 at 3:13 PM, Gregory Lee grle...@gmail.com wrote: I agree that often you don't need to worry about the memory order. However, it is not uncommon in medical imaging to go back and forth between a 2D or 3D image representation and a 1D array representation (e.g. as often used in image reconstruction algorithms). I found that the main time it was necessary to pay careful attention to the memory layout was when converting Matlab scripts that involve reshaping operations. Yes, good point.A typical example would be this kind of thing: # data is a 4D array with time / volume axis last data_2d = data.reshape((-1, data.shape[-1]) For MATLAB, the columns of this array would (by default) have the values on the first axis fastest changing, then the second, then the third, whereas numpy's default is the other way round. I find I usually don't have to worry about this, because I'm later going to do: data_processed_4d = data_2d.reshape(data.shape) which will reverse the previous reshape in the correct way. But in any case - this is not directly to do with the array memory layout. You will get the same output from reshape whether the memory layout of `data` was Fortran or C. Cheers, Matthew ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
On Mon Aug 3 16:26:10 2015 GMT+0200, Matthew Brett wrote: Hi, On Mon, Aug 3, 2015 at 3:13 PM, Gregory Lee grle...@gmail.com wrote: I agree that often you don't need to worry about the memory order. However, it is not uncommon in medical imaging to go back and forth between a 2D or 3D image representation and a 1D array representation (e.g. as often used in image reconstruction algorithms). I found that the main time it was necessary to pay careful attention to the memory layout was when converting Matlab scripts that involve reshaping operations. Yes, good point.A typical example would be this kind of thing: # data is a 4D array with time / volume axis last data_2d = data.reshape((-1, data.shape[-1]) For MATLAB, the columns of this array would (by default) have the values on the first axis fastest changing, then the second, then the third, whereas numpy's default is the other way round. I find I usually don't have to worry about this, because I'm later going to do: data_processed_4d = data_2d.reshape(data.shape) which will reverse the previous reshape in the correct way. But in any case - this is not directly to do with the array memory layout. You will get the same output from reshape whether the memory layout of `data` was Fortran or C. Just as a remark. Reshape has an (iteration not really memory) order parameter, thou it may do more copies if those do not match. - Sebastian Cheers, Matthew ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
Matthew Brett matthew.br...@gmail.com wrote: Sure, but to avoid confusion, maybe move the discussion of image indexing order to another thread? I think this thread is about memory layout, which is a different issue. It is actually a bit convoluted and not completely orthogonal. Memory layout does not matter for 2d ndexing, i.e. (x,y) vs. (row, column), if you are careful when iterating, but it does matter for Nd indexing. There is a reason to prefer (x,y,z,t,r) in column major order or (recording, time, slice, row, column) in row major order. Otherwise you can get very inefficient memory traversals. Then if you work with visualization libraries that expects (x,y,z) and column major order, e.g. ITK, VTK and OpenGL, this is really what you want to use. And the choise of indexing (x,y,z) cannot be seen as independent of the memory layout. Remember, it is not just a matter of mapping coordinates to pixels. The data sets are so large in MRI processing that memory layout does matter. Sturla ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
On Mon, Aug 3, 2015 at 5:01 PM, Sturla Molden sturla.mol...@gmail.com wrote: Matthew Brett matthew.br...@gmail.com wrote: Sure, but to avoid confusion, maybe move the discussion of image indexing order to another thread? I think this thread is about memory layout, which is a different issue. It is actually a bit convoluted and not completely orthogonal. Memory layout does not matter for 2d ndexing, i.e. (x,y) vs. (row, column), if you are careful when iterating, but it does matter for Nd indexing. There is a reason to prefer (x,y,z,t,r) in column major order or (recording, time, slice, row, column) in row major order. Otherwise you can get very inefficient memory traversals. Then if you work with visualization libraries that expects (x,y,z) and column major order, e.g. ITK, VTK and OpenGL, this is really what you want to use. And the choise of indexing (x,y,z) cannot be seen as independent of the memory layout. Remember, it is not just a matter of mapping coordinates to pixels. The data sets are so large in MRI processing that memory layout does matter. I completely agree that memory layout can affect performance, and this can be important. On the other hand, I think you agree that the relationship of axis order to memory layout is just a matter of mapping coordinates to pixels. So you can change the axis ordering without changing the memory layout and the memory layout without changing the axis ordering. Of course you could argue that it would be simpler to fuse the two issues, and enforce one memory layout - say Fortran.The result might well be easier think about, but it wouldn't be much like numpy, and it would have lots of performance and memory disadvantages. Cheers, Matthew ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
On Sun, Aug 2, 2015 at 1:46 PM, Sturla Molden sturla.mol...@gmail.com wrote: On 02/08/15 22:28, Bryan Van de Ven wrote: And to eliminate the order kwarg, use functools.partial to patch the zeros function (or any others, as needed): This will probably break code that depends on NumPy, like SciPy and scikit-image. But if NumPy is all that matters, sure go ahead and monkey patch. Otherwise keep the patched functions in another namespace. Id be really careful about this -- sure it's annoying, but a kind of global change of behavior could wreak havok. I'd create a set of Fortran-order constructors -- if it were me, I do: fzeros fones, etc. but you could, I suppose, create a namespace and ut hem all there, then create a fnumpy that would write those over: import numpy as np and away you go -- but that wouldn't change any code that imports numpy in the usual way. -Chris -- Christopher Barker, Ph.D. Oceanographer Emergency Response Division NOAA/NOS/ORR(206) 526-6959 voice 7600 Sand Point Way NE (206) 526-6329 fax Seattle, WA 98115 (206) 526-6317 main reception chris.bar...@noaa.gov ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
This is very good discussion. Thank you all for replying. I can see the fundamental difference is that I always think/talk/read/write a 3D image as I(x, y, z), not (plane, row, column) . I am coming from MRI (magnetic resonance imaging) research, and I can assure you that the entire MRI community is using (x, y, z), including books, journal papers, conference abstracts, presentations, everything. We even talk about what we called logical x/y/z and physical x/y/z, and the rotation matrix that converts the two coordinate systems. The radiologists are also used to (x, y, z). For example, we always say my image is 256 by 256 by 20 slices, and we never say 20 by 256 by 256. So, basically, at least in MRI, we always talk about an image as I(x, y, z), and we always assume that x is the fastest changing index. That's why I prefer column-major (because it is more natural). Of course, I can totally get my work done by using row-major, I just have to always remind myself write last dimension index first when coding. I actually have done this before, and I found it would be so much easier if just using column-major. Kang On 08/02/15, Juan Nunez-Iglesias jni.s...@gmail.com wrote: Hi Kang, Feel free to come chat about your application on the scikit-image list [1]! I'll note that we've been through the array order discussion many times there and even have a doc page about it [2]. The short version is that you'll save yourself a lot of pain by starting to think of your images as (plane, row, column) instead of (x, y, z). The syntax actually becomes friendlier too. For example, to do something to each slice of data, you do: for plane in image: plane += foo instead of for z in image.shape[2]: image[:, :, z] += foo for example. Juan. [1] scikit-im...@googlegroups.com [2] http://scikit-image.org/docs/dev/user_guide/numpy_images.html#coordinate-conventions PS: As to the renamed Fortran-ordered numpy, may I suggest funpy. The F is for Fortran and the fun is for all the fun you'll have maintaining it. =P On Mon, 3 Aug 2015 at 6:28 am Daniel Sank sank.dan...@gmail.com wrote: Kang, Thank you for explaining your motivation. It's clear from your last note, as you said, that your desire for column-first indexing has nothing to do with in-memory data layout. That being the case, I strongly urge you to just use bare numpy and do not use the fortran_zeros function I recommended before. Changing the in-memory layout via the order keyword in numpy.zeros will not change the way indexing works at all. You gain absolutely nothing by changing the in-memory order unless you are writing some C or Fortran code which will interact with the data in memory. To see what I mean, consider the following examples: x = np.array([1, 2, 3], [4, 5, 6]]) x.shape (2, 3) and x = np.array([1, 2, 3], [4, 5, 6]], order='F') x.shape (2, 3) You see that changing the in-memory order has nothing whatsoever to do with the array's shape or how you access it. You will see run time error. Depending on environment, you may get useful error message (i.e. index out of range), but sometimes you just get bad image results. Could you give a very simple example of what you mean? I can't think of how this could ever happen and your fear here makes me think there's a fundamental misunderstanding about how array operations in numpy and other programming languages work. As an example, iteration in numpy goes through the first index: x = np.array([[1, 2, 3], [4, 5, 6]]) for foo in x: ... Inside the for loop, foo takes on the values [1, 2, 3] on the first iteration and [4, 5, 6] on the second. If you want to iterate through the columns just do this instead x = np.array([[1, 2, 3], [4, 5, 6]]) for foo in x.T: ... If your complaint is that you want np.array([[1, 2, 3], [4, 5, 6]]) to produce an array with shape (3, 2) then you should own up to the fact that the array constructor expects it the other way around and do this x = np.array([[1, 2, 3], [4, 5, 6]]).T instead. This is infinity times better than trying to write a shim function or patch numpy because with .T you're using (fast) built-in functionality which other people your code will understand. The real message here is that whether the first index runs over rows or columns is actually meaningless. The only places the row versus column issue has any meaning is when doing input/output (in which case you should use the transpose if you actually need it), or when doing iteration. One thing that would make sense if you're reading from a binary file format which uses column-major format would be to write your own reader function: def read_fortran_style_binary_file(file): return np.fromfile(file).T
Re: [Numpy-discussion] Change default order to Fortran order
On Aug 2, 2015 11:06 PM, Kang Wang kwan...@wisc.edu wrote: This is very good discussion. Thank you all for replying. I can see the fundamental difference is that I always think/talk/read/write a 3D image as I(x, y, z), not (plane, row, column) . I am coming from MRI (magnetic resonance imaging) research, and I can assure you that the entire MRI community is using (x, y, z), including books, journal papers, conference abstracts, presentations, everything. We even talk about what we called logical x/y/z and physical x/y/z, and the rotation matrix that converts the two coordinate systems. The radiologists are also used to (x, y, z). For example, we always say my image is 256 by 256 by 20 slices, and we never say 20 by 256 by 256. So, basically, at least in MRI, we always talk about an image as I(x, y, z), and we always assume that x is the fastest changing index. That's why I prefer column-major (because it is more natural). Of course, I can totally get my work done by using row-major, I just have to always remind myself write last dimension index first when coding. I actually have done this before, and I found it would be so much easier if just using column-major. Why not just use I[x, y, z] like you're used to, and let the computer worry about the physical layout in memory? Sometimes this will be Fortran order and sometimes C order and sometimes something else, but you don't have to know or care; 99% of the time it doesn't matter. The worst case is that when you use a python wrapper to call into a library that can only handle Fortran order, then the wrapper will quietly have to convert the memory order around and it will be slightly slower than if you had used Fortran order in the first place. But in practice you'll barely ever notice this, and when you do, *then* you can tell numpy explicitly what memory layout you want in the situation where it matters. General principle: do what's easiest for the programmer, not what's easiest for the computer. -n ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
Hi,
On Mon, Aug 3, 2015 at 8:09 AM, Nathaniel Smith n...@pobox.com wrote:
On Aug 2, 2015 11:06 PM, Kang Wang kwan...@wisc.edu wrote:
This is very good discussion. Thank you all for replying.
I can see the fundamental difference is that I always
think/talk/read/write a 3D image as I(x, y, z), not (plane, row, column) . I
am coming from MRI (magnetic resonance imaging) research, and I can assure
you that the entire MRI community is using (x, y, z), including books,
journal papers, conference abstracts, presentations, everything. We even
talk about what we called logical x/y/z and physical x/y/z, and the
rotation matrix that converts the two coordinate systems. The radiologists
are also used to (x, y, z). For example, we always say my image is 256 by
256 by 20 slices, and we never say 20 by 256 by 256.
So, basically, at least in MRI, we always talk about an image as I(x, y,
z), and we always assume that x is the fastest changing index. That's why
I prefer column-major (because it is more natural).
Of course, I can totally get my work done by using row-major, I just have
to always remind myself write last dimension index first when coding. I
actually have done this before, and I found it would be so much easier if
just using column-major.
Why not just use I[x, y, z] like you're used to, and let the computer worry
about the physical layout in memory? Sometimes this will be Fortran order
and sometimes C order and sometimes something else, but you don't have to
know or care; 99% of the time it doesn't matter. The worst case is that when
you use a python wrapper to call into a library that can only handle Fortran
order, then the wrapper will quietly have to convert the memory order around
and it will be slightly slower than if you had used Fortran order in the
first place. But in practice you'll barely ever notice this, and when you
do, *then* you can tell numpy explicitly what memory layout you want in the
situation where it matters.
Yes - if you are using numpy, you really have to look numpy in the eye and say:
I will let you worry about the array element order in memory, and in
return, you promise to make indexing work as I would expect
Just for example, let's say you loaded an MRI image into memory:
In [1]: import nibabel
In [2]: img = nibabel.load('my_mri.nii')
In [3]: data = img.get_data()
Because NIfTI images are Fortran memory layout, this happens to be the
memory layout you get for your array:
In [4]: data.flags
Out[4]:
C_CONTIGUOUS : False
F_CONTIGUOUS : True
OWNDATA : False
WRITEABLE : True
ALIGNED : True
UPDATEIFCOPY : False
But now - in Python - all I care about is what data I have on the
first, second, third axes. For example, I could do this:
In [5]: data_copy = data.copy()
This has exactly the same values as the original array, and at the
same index positions:
In [7]: import numpy as np
In [8]: np.all(data == data)
Out[8]: memmap(True, dtype=bool)
but I now have a C memory layout array.
In [9]: data_copy.flags
Out[9]:
C_CONTIGUOUS : True
F_CONTIGUOUS : False
OWNDATA : True
WRITEABLE : True
ALIGNED : True
UPDATEIFCOPY : False
Worse than that, if I slice my original data array, then I get an
array that is neither C- or Fortran- compatible in memory:
In [10]: data_view = data[:, :, ::2]
In [11]: data_view.flags
Out[11]:
C_CONTIGUOUS : False
F_CONTIGUOUS : False
OWNDATA : False
WRITEABLE : True
ALIGNED : True
UPDATEIFCOPY : False
So - if you want every array to be Fortran-contiguous in memory, I
would not start with numpy at all, I would write your own array
library.
The alternative - or the numpy way - is to give up on enforcing a
particular layout in memory, until you need to pass an array to some C
or C++ or Fortran code that needs some particular layout, in which
case you get your extension code to copy the array into the required
layout on entry. Of course this is what numpy itself has to do when
interfacing with external libraries like BLAS or LAPACK.
Cheers,
Matthew
___
NumPy-Discussion mailing list
NumPy-Discussion@scipy.org
http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
Juan Nunez-Iglesias jni.s...@gmail.com wrote: The short version is that you'll save yourself a lot of pain by starting to think of your images as (plane, row, column) instead of (x, y, z). There are several things to consider here. 1. The vertices in computer graphics (OpenGL) are (x,y,z). 2. OpenGL rotation matrices and projection matrice are stored in column major order. 3. OpenGL frame buffers are indexed (x,y) in column major order with (0,0) being lower left. 4. ITK and VTK depends on OpenGL and are thus using column major order. 5. Those who use Matlab or Fortran in addition to Python prefer column major order. 6. BLAS and LAPACK use column major order. 7. The common notation in image prorcessing (as opposed to computer graphics in geberal) is indexing (row, column), in row major order, with (0,0) being upper left. All in all, this is a strong case for prefering column major order and the common mathematical notation (x,y,z). Also notice how the ususal notation in image pricessing differs from OpenGL. Sturla ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
well its really great idea. i can help on python but i don't have any knowledge on fortran. On Sun, Aug 2, 2015 at 7:25 PM, Kang Wang kwan...@wisc.edu wrote: Hi, I am an imaging researcher, and a new Python user. My first Python project is to somehow modify NumPy source code such that everything is Fortran column-major by default. I read about the information in the link below, but for us, the fact is that *we absolutely want to use Fortran column major, and we want to make it default. Explicitly writing order = 'F' all over the place is not acceptable to us*. http://docs.scipy.org/doc/numpy/reference/internals.html#multidimensional-array-indexing-order-issues I tried searching in this email list, as well as google search in general. However, I have not found anything useful. This must be a common request/need, I believe. Can anyone provide any insight/help? Thank you very much, Kang -- *Kang Wang, Ph.D.* Highland Ave., Room 1113 Madison, WI 53705-2275 ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
On Mon, Aug 3, 2015 at 10:24 AM, Matthew Brett matthew.br...@gmail.com wrote: On Mon, Aug 3, 2015 at 5:01 PM, Sturla Molden sturla.mol...@gmail.com wrote: Matthew Brett matthew.br...@gmail.com wrote: Sure, but to avoid confusion, maybe move the discussion of image indexing order to another thread? I think this thread is about memory layout, which is a different issue. It is actually a bit convoluted and not completely orthogonal. Memory layout does not matter for 2d ndexing, i.e. (x,y) vs. (row, column), if you are careful when iterating, but it does matter for Nd indexing. There is a reason to prefer (x,y,z,t,r) in column major order or (recording, time, slice, row, column) in row major order. Otherwise you can get very inefficient memory traversals. Then if you work with visualization libraries that expects (x,y,z) and column major order, e.g. ITK, VTK and OpenGL, this is really what you want to use. And the choise of indexing (x,y,z) cannot be seen as independent of the memory layout. Remember, it is not just a matter of mapping coordinates to pixels. The data sets are so large in MRI processing that memory layout does matter. I completely agree that memory layout can affect performance, and this can be important. On the other hand, I think you agree that the relationship of axis order to memory layout is just a matter of mapping coordinates to pixels. So you can change the axis ordering without changing the memory layout and the memory layout without changing the axis ordering. Of course you could argue that it would be simpler to fuse the two issues, and enforce one memory layout - say Fortran.The result might well be easier think about, but it wouldn't be much like numpy, and it would have lots of performance and memory disadvantages. Cheers, I would also strongly suggest that once you have decided on a convention that you thoroughly document it somewhere. That will not only help you, but anyone who later needs to maintain the code will bless you rather than d*nm you to eternal torture by the seven demons of Ipsos. Chuck ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
On Mon Aug 3 10:49:35 2015 GMT+0200, Matthew Brett wrote:
Hi,
On Mon, Aug 3, 2015 at 8:09 AM, Nathaniel Smith n...@pobox.com wrote:
On Aug 2, 2015 11:06 PM, Kang Wang kwan...@wisc.edu wrote:
This is very good discussion. Thank you all for replying.
I can see the fundamental difference is that I always
think/talk/read/write a 3D image as I(x, y, z), not (plane, row, column) .
I
am coming from MRI (magnetic resonance imaging) research, and I can assure
you that the entire MRI community is using (x, y, z), including books,
journal papers, conference abstracts, presentations, everything. We even
talk about what we called logical x/y/z and physical x/y/z, and the
rotation matrix that converts the two coordinate systems. The radiologists
are also used to (x, y, z). For example, we always say my image is 256 by
256 by 20 slices, and we never say 20 by 256 by 256.
So, basically, at least in MRI, we always talk about an image as I(x, y,
z), and we always assume that x is the fastest changing index. That's why
I prefer column-major (because it is more natural).
Of course, I can totally get my work done by using row-major, I just have
to always remind myself write last dimension index first when coding. I
actually have done this before, and I found it would be so much easier if
just using column-major.
Why not just use I[x, y, z] like you're used to, and let the computer worry
about the physical layout in memory? Sometimes this will be Fortran order
and sometimes C order and sometimes something else, but you don't have to
know or care; 99% of the time it doesn't matter. The worst case is that when
you use a python wrapper to call into a library that can only handle Fortran
order, then the wrapper will quietly have to convert the memory order around
and it will be slightly slower than if you had used Fortran order in the
first place. But in practice you'll barely ever notice this, and when you
do, *then* you can tell numpy explicitly what memory layout you want in the
situation where it matters.
Yes - if you are using numpy, you really have to look numpy in the eye and
say:
I will let you worry about the array element order in memory, and in
return, you promise to make indexing work as I would expect
Just for example, let's say you loaded an MRI image into memory:
In [1]: import nibabel
In [2]: img = nibabel.load('my_mri.nii')
In [3]: data = img.get_data()
Because NIfTI images are Fortran memory layout, this happens to be the
memory layout you get for your array:
In [4]: data.flags
Out[4]:
C_CONTIGUOUS : False
F_CONTIGUOUS : True
OWNDATA : False
WRITEABLE : True
ALIGNED : True
UPDATEIFCOPY : False
But now - in Python - all I care about is what data I have on the
first, second, third axes. For example, I could do this:
In [5]: data_copy = data.copy()
This has exactly the same values as the original array, and at the
same index positions:
In [7]: import numpy as np
In [8]: np.all(data == data)
Out[8]: memmap(True, dtype=bool)
but I now have a C memory layout array.
In [9]: data_copy.flags
Out[9]:
C_CONTIGUOUS : True
F_CONTIGUOUS : False
OWNDATA : True
WRITEABLE : True
ALIGNED : True
UPDATEIFCOPY : False
Yeah, I would like to second those arguments. Most of the time, there is no
need to worry about layout. For large chunks you allocate, it may make sense
for speed, etc. So you can alias creation functions. Generally, I would
suggest to simply not worry about the memory layout. Also do not *trust* the
layout for most function returns. If you need a specific layout to interface
other code, always check what you got it.
-Sebastian
Worse than that, if I slice my original data array, then I get an
array that is neither C- or Fortran- compatible in memory:
In [10]: data_view = data[:, :, ::2]
In [11]: data_view.flags
Out[11]:
C_CONTIGUOUS : False
F_CONTIGUOUS : False
OWNDATA : False
WRITEABLE : True
ALIGNED : True
UPDATEIFCOPY : False
So - if you want every array to be Fortran-contiguous in memory, I
would not start with numpy at all, I would write your own array
library.
The alternative - or the numpy way - is to give up on enforcing a
particular layout in memory, until you need to pass an array to some C
or C++ or Fortran code that needs some particular layout, in which
case you get your extension code to copy the array into the required
layout on entry. Of course this is what numpy itself has to do when
interfacing with external libraries like BLAS or LAPACK.
Cheers,
Matthew
___
NumPy-Discussion mailing list
NumPy-Discussion@scipy.org
http://mail.scipy.org/mailman/listinfo/numpy-discussion
___
NumPy-Discussion mailing list
NumPy-Discussion@scipy.org
http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
I agree that often you don't need to worry about the memory order.
However, it is not uncommon in medical imaging to go back and forth between
a 2D or 3D image representation and a 1D array representation (e.g. as
often used in image reconstruction algorithms). I found that the main time
it was necessary to pay careful attention to the memory layout was when
converting Matlab scripts that involve reshaping operations.
On Mon, Aug 3, 2015 at 8:02 AM, Sebastian Berg sebast...@sipsolutions.net
wrote:
On Mon Aug 3 10:49:35 2015 GMT+0200, Matthew Brett wrote:
Hi,
On Mon, Aug 3, 2015 at 8:09 AM, Nathaniel Smith n...@pobox.com wrote:
On Aug 2, 2015 11:06 PM, Kang Wang kwan...@wisc.edu wrote:
This is very good discussion. Thank you all for replying.
I can see the fundamental difference is that I always
think/talk/read/write a 3D image as I(x, y, z), not (plane, row,
column) . I
am coming from MRI (magnetic resonance imaging) research, and I can
assure
you that the entire MRI community is using (x, y, z), including books,
journal papers, conference abstracts, presentations, everything. We
even
talk about what we called logical x/y/z and physical x/y/z, and
the
rotation matrix that converts the two coordinate systems. The
radiologists
are also used to (x, y, z). For example, we always say my image is
256 by
256 by 20 slices, and we never say 20 by 256 by 256.
So, basically, at least in MRI, we always talk about an image as I(x,
y,
z), and we always assume that x is the fastest changing index.
That's why
I prefer column-major (because it is more natural).
Of course, I can totally get my work done by using row-major, I just
have
to always remind myself write last dimension index first when
coding. I
actually have done this before, and I found it would be so much
easier if
just using column-major.
Why not just use I[x, y, z] like you're used to, and let the computer
worry
about the physical layout in memory? Sometimes this will be Fortran
order
and sometimes C order and sometimes something else, but you don't have
to
know or care; 99% of the time it doesn't matter. The worst case is
that when
you use a python wrapper to call into a library that can only handle
Fortran
order, then the wrapper will quietly have to convert the memory order
around
and it will be slightly slower than if you had used Fortran order in
the
first place. But in practice you'll barely ever notice this, and when
you
do, *then* you can tell numpy explicitly what memory layout you want
in the
situation where it matters.
Yes - if you are using numpy, you really have to look numpy in the eye
and say:
I will let you worry about the array element order in memory, and in
return, you promise to make indexing work as I would expect
Just for example, let's say you loaded an MRI image into memory:
In [1]: import nibabel
In [2]: img = nibabel.load('my_mri.nii')
In [3]: data = img.get_data()
Because NIfTI images are Fortran memory layout, this happens to be the
memory layout you get for your array:
In [4]: data.flags
Out[4]:
C_CONTIGUOUS : False
F_CONTIGUOUS : True
OWNDATA : False
WRITEABLE : True
ALIGNED : True
UPDATEIFCOPY : False
But now - in Python - all I care about is what data I have on the
first, second, third axes. For example, I could do this:
In [5]: data_copy = data.copy()
This has exactly the same values as the original array, and at the
same index positions:
In [7]: import numpy as np
In [8]: np.all(data == data)
Out[8]: memmap(True, dtype=bool)
but I now have a C memory layout array.
In [9]: data_copy.flags
Out[9]:
C_CONTIGUOUS : True
F_CONTIGUOUS : False
OWNDATA : True
WRITEABLE : True
ALIGNED : True
UPDATEIFCOPY : False
Yeah, I would like to second those arguments. Most of the time, there is
no need to worry about layout. For large chunks you allocate, it may make
sense for speed, etc. So you can alias creation functions. Generally, I
would suggest to simply not worry about the memory layout. Also do not
*trust* the layout for most function returns. If you need a specific layout
to interface other code, always check what you got it.
-Sebastian
Worse than that, if I slice my original data array, then I get an
array that is neither C- or Fortran- compatible in memory:
In [10]: data_view = data[:, :, ::2]
In [11]: data_view.flags
Out[11]:
C_CONTIGUOUS : False
F_CONTIGUOUS : False
OWNDATA : False
WRITEABLE : True
ALIGNED : True
UPDATEIFCOPY : False
So - if you want every array to be Fortran-contiguous in memory, I
would not start with numpy at all, I would write your own array
library.
The alternative - or the numpy way - is to give up on enforcing a
particular layout in memory, until you need to pass an array to some C
or C++ or Fortran
Re: [Numpy-discussion] Change default order to Fortran order
Kang, Thank you for explaining your motivation. It's clear from your last note, as you said, that your desire for column-first indexing has nothing to do with in-memory data layout. That being the case, I strongly urge you to just use bare numpy and do not use the fortran_zeros function I recommended before. Changing the in-memory layout via the order keyword in numpy.zeros will not change the way indexing works at all. You gain absolutely nothing by changing the in-memory order unless you are writing some C or Fortran code which will interact with the data in memory. To see what I mean, consider the following examples: x = np.array([1, 2, 3], [4, 5, 6]]) x.shape (2, 3) and x = np.array([1, 2, 3], [4, 5, 6]], order='F') x.shape (2, 3) You see that changing the in-memory order has nothing whatsoever to do with the array's shape or how you access it. You will see run time error. Depending on environment, you may get useful error message (i.e. index out of range), but sometimes you just get bad image results. Could you give a very simple example of what you mean? I can't think of how this could ever happen and your fear here makes me think there's a fundamental misunderstanding about how array operations in numpy and other programming languages work. As an example, iteration in numpy goes through the first index: x = np.array([[1, 2, 3], [4, 5, 6]]) for foo in x: ... Inside the for loop, foo takes on the values [1, 2, 3] on the first iteration and [4, 5, 6] on the second. If you want to iterate through the columns just do this instead x = np.array([[1, 2, 3], [4, 5, 6]]) for foo in x.T: ... If your complaint is that you want np.array([[1, 2, 3], [4, 5, 6]]) to produce an array with shape (3, 2) then you should own up to the fact that the array constructor expects it the other way around and do this x = np.array([[1, 2, 3], [4, 5, 6]]).T instead. This is infinity times better than trying to write a shim function or patch numpy because with .T you're using (fast) built-in functionality which other people your code will understand. The real message here is that whether the first index runs over rows or columns is actually meaningless. The only places the row versus column issue has any meaning is when doing input/output (in which case you should use the transpose if you actually need it), or when doing iteration. One thing that would make sense if you're reading from a binary file format which uses column-major format would be to write your own reader function: def read_fortran_style_binary_file(file): return np.fromfile(file).T Note that if you do this then you already have a column major array in numpy and you don't have to worry about any other transposes (except, again, when doing more I/O or passing to something like a plotting function). On Sun, Aug 2, 2015 at 7:16 PM, Kang Wang kwan...@wisc.edu wrote: Thank you all for replying and providing useful insights and suggestions. The reasons I really want to use column-major are: - I am image-oriented user (not matrix-oriented, as explained in http://docs.scipy.org/doc/numpy/reference/internals.html#multidimensional-array-indexing-order-issues ) - I am so used to read/write I(x, y, z) in textbook and code, and it is very likely that if the environment (row-major environment) forces me to write I(z, y, x), I will write a bug if I am not 100% focused. When this happens, it is difficult to debug, because everything compile and build fine. You will see run time error. Depending on environment, you may get useful error message (i.e. index out of range), but sometimes you just get bad image results. - It actually has not too much to do with the actual data layout in memory. In imaging processing, especially medical imaging where I am working in, if you have a 3D image, everyone will agree that in memory, the X index is the fasted changing index, and the Z dimension (we often call it the slice dimension) has the largest stride in memory. So, if data layout is like this in memory, and image-oriented users are so used to read/write I(x,y,z), the only storage order that makes sense is column-major - I also write code in MATLAB and C/C++. In MATLAB, matrix is column-major array. In C/C++, we often use ITK, which is also column-major ( http://www.itk.org/Doxygen/html/classitk_1_1Image.html). I really prefer always read/write column-major code to minimize coding bugs related to storage order. - I also prefer index to be 0-based; however, there is nothing I can do about it for MATLAB (which is 1-based). I can see that my original thought about modifying NumPy source and re-compile is probably a bad idea. The suggestions about using fortran_zeros = partial(np.zeros(order='F')) is probably the best way so far, in my opinion, and I am going to give it a try. Again, thank you all for replying. Kang On 08/02/15, *Nathaniel
Re: [Numpy-discussion] Change default order to Fortran order
Hi Kang, Feel free to come chat about your application on the scikit-image list [1]! I'll note that we've been through the array order discussion many times there and even have a doc page about it [2]. The short version is that you'll save yourself a lot of pain by starting to think of your images as (plane, row, column) instead of (x, y, z). The syntax actually becomes friendlier too. For example, to do something to each slice of data, you do: for plane in image: plane += foo instead of for z in image.shape[2]: image[:, :, z] += foo for example. Juan. [1] scikit-im...@googlegroups.com [2] http://scikit-image.org/docs/dev/user_guide/numpy_images.html#coordinate-conventions PS: As to the renamed Fortran-ordered numpy, may I suggest funpy. The F is for Fortran and the fun is for all the fun you'll have maintaining it. =P On Mon, 3 Aug 2015 at 6:28 am Daniel Sank sank.dan...@gmail.com wrote: Kang, Thank you for explaining your motivation. It's clear from your last note, as you said, that your desire for column-first indexing has nothing to do with in-memory data layout. That being the case, I strongly urge you to just use bare numpy and do not use the fortran_zeros function I recommended before. Changing the in-memory layout via the order keyword in numpy.zeros will not change the way indexing works at all. You gain absolutely nothing by changing the in-memory order unless you are writing some C or Fortran code which will interact with the data in memory. To see what I mean, consider the following examples: x = np.array([1, 2, 3], [4, 5, 6]]) x.shape (2, 3) and x = np.array([1, 2, 3], [4, 5, 6]], order='F') x.shape (2, 3) You see that changing the in-memory order has nothing whatsoever to do with the array's shape or how you access it. You will see run time error. Depending on environment, you may get useful error message (i.e. index out of range), but sometimes you just get bad image results. Could you give a very simple example of what you mean? I can't think of how this could ever happen and your fear here makes me think there's a fundamental misunderstanding about how array operations in numpy and other programming languages work. As an example, iteration in numpy goes through the first index: x = np.array([[1, 2, 3], [4, 5, 6]]) for foo in x: ... Inside the for loop, foo takes on the values [1, 2, 3] on the first iteration and [4, 5, 6] on the second. If you want to iterate through the columns just do this instead x = np.array([[1, 2, 3], [4, 5, 6]]) for foo in x.T: ... If your complaint is that you want np.array([[1, 2, 3], [4, 5, 6]]) to produce an array with shape (3, 2) then you should own up to the fact that the array constructor expects it the other way around and do this x = np.array([[1, 2, 3], [4, 5, 6]]).T instead. This is infinity times better than trying to write a shim function or patch numpy because with .T you're using (fast) built-in functionality which other people your code will understand. The real message here is that whether the first index runs over rows or columns is actually meaningless. The only places the row versus column issue has any meaning is when doing input/output (in which case you should use the transpose if you actually need it), or when doing iteration. One thing that would make sense if you're reading from a binary file format which uses column-major format would be to write your own reader function: def read_fortran_style_binary_file(file): return np.fromfile(file).T Note that if you do this then you already have a column major array in numpy and you don't have to worry about any other transposes (except, again, when doing more I/O or passing to something like a plotting function). On Sun, Aug 2, 2015 at 7:16 PM, Kang Wang kwan...@wisc.edu wrote: Thank you all for replying and providing useful insights and suggestions. The reasons I really want to use column-major are: - I am image-oriented user (not matrix-oriented, as explained in http://docs.scipy.org/doc/numpy/reference/internals.html#multidimensional-array-indexing-order-issues ) - I am so used to read/write I(x, y, z) in textbook and code, and it is very likely that if the environment (row-major environment) forces me to write I(z, y, x), I will write a bug if I am not 100% focused. When this happens, it is difficult to debug, because everything compile and build fine. You will see run time error. Depending on environment, you may get useful error message (i.e. index out of range), but sometimes you just get bad image results. - It actually has not too much to do with the actual data layout in memory. In imaging processing, especially medical imaging where I am working in, if you have a 3D image, everyone will agree that in memory, the X index is the fasted changing index, and the Z dimension (we often call it the slice
[Numpy-discussion] Change default order to Fortran order
Hi, I am an imaging researcher, and a new Python user. My first Python project is to somehow modify NumPy source code such that everything is Fortran column-major by default. I read about the information in the link below, but for us, the fact is that we absolutely want to use Fortran column major, and we want to make it default. Explicitly writing order = 'F' all over the place is not acceptable to us. http://docs.scipy.org/doc/numpy/reference/internals.html#multidimensional-array-indexing-order-issues I tried searching in this email list, as well as google search in general. However, I have not found anything useful. This must be a common request/need, I believe. Can anyone provide any insight/help? Thank you very much, Kang -- Kang Wang, Ph.D. Highland Ave., Room 1113 Madison, WI 53705-2275 ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
On 02/08/15 15:55, Kang Wang wrote: Can anyone provide any insight/help? There is no default order. There was before, but now all operators control the order of their return arrays from the order of their input array. The only thing that makes C order default is the keyword argument to np.empty, np.ones and np.zeros. Just monkey patch those functions and it should be fine. Sturla ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
Well, numpy has a tendency to prefer C order. There is nothing you can do about that really. But you just cannot be sure what you get in some cases. Often you need something specific for interfaceing other code. But in that case quite often you also do not need to fear the copy. - Sebastian On Sun Aug 2 16:27:08 2015 GMT+0200, Sturla Molden wrote: On 02/08/15 15:55, Kang Wang wrote: Can anyone provide any insight/help? There is no default order. There was before, but now all operators control the order of their return arrays from the order of their input array. The only thing that makes C order default is the keyword argument to np.empty, np.ones and np.zeros. Just monkey patch those functions and it should be fine. Sturla ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
On 02/08/15 22:28, Bryan Van de Ven wrote: And to eliminate the order kwarg, use functools.partial to patch the zeros function (or any others, as needed): This will probably break code that depends on NumPy, like SciPy and scikit-image. But if NumPy is all that matters, sure go ahead and monkey patch. Otherwise keep the patched functions in another namespace. :-) Sturla ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
Could you please explain why you need 'F' ordering? It's pretty unlikely that you actually care about the internal memory layout, and you'll get better advice if you explain why you think you do care. My first Python project is to somehow modify NumPy source code such that everything is Fortran column-major by default. This is the road to pain. You'll have to maintain your own fork and will probably inject bugs when trying to rewrite. Nobody will want to help fix them because everyone else just uses numpy as is. And to eliminate the order kwarg, use functools.partial to patch the zeros function (or any others, as needed): Instead of monkey patching, why not just define your own shims: fortran_zeros = partial(np.zeros(order='F')) Seems like this would lead to a lot less confusion (although until you tell us why you care about the in-memory layout I don't know the point of doing this at all). ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
On Aug 2, 2015 7:30 AM, Sturla Molden sturla.mol...@gmail.com wrote: On 02/08/15 15:55, Kang Wang wrote: Can anyone provide any insight/help? There is no default order. There was before, but now all operators control the order of their return arrays from the order of their input array. This is... overoptimistic. I would not rely on this in code that I wrote. It's true that many numpy operations do preserve the input order. But there are also many operations that don't. And which are which often changes between releases. (Not on purpose usually, but it's an easy bug to introduce. And sometimes it is done intentionally, e.g. to make functions faster. It sucks to have to make a function slower for everyone because someone somewhere is depending on memory layout default details.) And there are operations where it's not even clear what preserving order means (indexing a C array with a Fortran array, add(C, fortran), ...), and even lots of operations that intrinsically break contiguity/ordering (transpose, diagonal, slicing, swapaxes, ...), so you will end up with mixed orderings one way or another in any non-trivial program. Instead, it's better to explicitly specify order= just at the places where you care. That way your code is *locally* correct (you can check it will work by just reading the one function). The alternative is to try and enforce a *global* property on your program (everyone everywhere is very careful to only use contiguity-preserving operations, where everyone includes third party libraries like numpy and others). In software design, local invariants invariants are always better than global invariants -- the most well known example is local variables versus global variables, but the principle is much broader. -n ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
On 02/08/15 22:14, Kang Wang wrote: Thank you all for replying! I did a quick test, using python 2.6.6, and the original numpy package on my Linux computer without any change. == x = np.zeros((2,3),dtype=np.int32,order='F') print x.strides = print x.strides y = x + 1 print y.strides = print y.strides == Output: x.strides = (4, 8) y.strides = (12, 4) Update NumPy. This is the behavior I talked about that has changed. Now NumPy does this: In [21]: x = np.zeros((2,3),dtype=np.int32,order='F') In [22]: y = x + 1 In [24]: x.strides Out[24]: (4, 8) In [25]: y.strides Out[25]: (4, 8) Sturla ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
Thank you all for replying! I did a quick test, using python 2.6.6, and the original numpy package on my Linux computer without any change.== x = np.zeros((2,3),dtype=np.int32,order='F') print x.strides = print x.strides y = x + 1 print y.strides = print y.strides == Output: x.strides = (4, 8) y.strides = (12, 4) So, basically, x is Fortran-style column-major (because I explicitly write order='F'), but y is C-style row-major. This is going to be very annoying. What I really want is: - I do not have to write order='F' explicitly when declaring x - both x and y are Fortran-style column-major Which file should I modify to achieve this goal? Right now, I am just trying to get some basic stuff working with all arrays default to Fortran-style, and I can worry about interfacing with other code/libraries later. Thanks, Kang On 08/02/15, Sebastian Berg sebast...@sipsolutions.net wrote: Well, numpy has a tendency to prefer C order. There is nothing you can do about that really. But you just cannot be sure what you get in some cases. Often you need something specific for interfaceing other code. But in that case quite often you also do not need to fear the copy. - Sebastian On Sun Aug 2 16:27:08 2015 GMT+0200, Sturla Molden wrote: On 02/08/15 15:55, Kang Wang wrote: Can anyone provide any insight/help? There is no default order. There was before, but now all operators control the order of their return arrays from the order of their input array. The only thing that makes C order default is the keyword argument to np.empty, np.ones and np.zeros. Just monkey patch those functions and it should be fine. Sturla ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion -- Kang Wang, Ph.D. Highland Ave., Room 1113 Madison, WI 53705-2275 TEL 608-263-0066 http://www.medphysics.wisc.edu/~kang/ ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
And to eliminate the order kwarg, use functools.partial to patch the zeros function (or any others, as needed): In [26]: import numpy as np In [27]: from functools import partial In [28]: np.zeros = partial(np.zeros, order=F) In [29]: x = np.zeros((2,3), dtype=np.int32) In [30]: y = x + 1 In [31]: x.strides Out[31]: (4, 8) In [32]: y.strides Out[32]: (4, 8) In [33]: np.__version__ Out[33]: '1.9.2' Bryan On Aug 2, 2015, at 3:22 PM, Sturla Molden sturla.mol...@gmail.com wrote: On 02/08/15 22:14, Kang Wang wrote: Thank you all for replying! I did a quick test, using python 2.6.6, and the original numpy package on my Linux computer without any change. == x = np.zeros((2,3),dtype=np.int32,order='F') print x.strides = print x.strides y = x + 1 print y.strides = print y.strides == Output: x.strides = (4, 8) y.strides = (12, 4) Update NumPy. This is the behavior I talked about that has changed. Now NumPy does this: In [21]: x = np.zeros((2,3),dtype=np.int32,order='F') In [22]: y = x + 1 In [24]: x.strides Out[24]: (4, 8) In [25]: y.strides Out[25]: (4, 8) Sturla ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
On Aug 2, 2015 6:59 AM, Kang Wang kwan...@wisc.edu wrote:
Hi,
I am an imaging researcher, and a new Python user. My first Python
project is to somehow modify NumPy source code such that everything is
Fortran column-major by default.
I read about the information in the link below, but for us, the fact is
that we absolutely want to use Fortran column major, and we want to make
it default. Explicitly writing order = 'F' all over the place is not
acceptable to us.
http://docs.scipy.org/doc/numpy/reference/internals.html#multidimensional-array-indexing-order-issues
I tried searching in this email list, as well as google search in
general. However, I have not found anything useful. This must be a common
request/need, I believe.
It isn't, I'm afraid. Basically what you're signing up for is to maintain
your own copy of numpy all by yourself. You're totally within your rights
to do this, but it isn't something I would really recommend as a first
python project (to put it mildly).
And unfortunately, there are plenty of libraries out there that use numpy
and assume they will get C order by default, so your version of numpy will
create lots of obscure errors, segfaults, etc. as you start using it with
other packages. Obviously this will be a problem for you -- basically you
may find yourself having to maintain your own copy of lots of libraries.
Less obviously, this would also create a big problem for us, because your
users will start filling bug reports on numpy, or on these random third
party packages, and it will be massively confusing and a big waste of time
because the problem will be with your package, not with any of our code. So
if you do do this, please either (a) change the name of your package
somehow ('import numpyfortran' or similar) so that everyone using it is
clear that it's a non-standard product, or else (b) make sure that you only
use it within your own team, don't allow anyone else to use it, and make a
rule that no one is allowed to file bug reports, or ask or answer questions
on mailing lists or stackoverflow, unless they have first double checked
*every* time that what they're saying is also valid when using regular
numpy.
Again, I strongly recommend you not do this.
There are literally millions of users who are using numpy as it currently
is, and able to get stuff done. I don't know your specific situation, but
maybe if you describe a bit more what it is you're doing and why you think
you need all-Fortran-all-the-time, then people will be able to suggest
strategies to work around things on your end, or find smaller tweaks to
numpy that could go into the standard version.
-n
___
NumPy-Discussion mailing list
NumPy-Discussion@scipy.org
http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
On Aug 2, 2015 1:17 PM, Kang Wang kwan...@wisc.edu wrote: Thank you all for replying! I did a quick test, using python 2.6.6, There's pretty much no good reason these days to be using python 2.6 (which was released in *2008*). I assume you're using it because you're using redhat or some redhat derivative, and that's what they ship by default? Even redhat engineers officially recommend that users *not* use the default python -- it's basically only intended for use by their own built-in system management scripts. If you're just getting started with python, then at this point I'd recommend starting with python 3.4. Some easy ways to get this installed: - Anaconda: the most popular scientific python distribution -- you pretty much just download one file and get a full, up to date setup of python and all the main scientific packages, in your home directory. Supported on all popular platforms. Trivial to use and requires no special permissions. http://continuum.io/downloads#py34 - One of Anaconda's competitors: http://www.scipy.org/install.html - Software collections: redhat's official way to do things like this: https://www.softwarecollections.org/en/ -n ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Change default order to Fortran order
Just chiming in with my 2 cents, in direct response to your points... - Image oriented processing is most typically done with row-major storage layout. From hardware to general software implementations. - Well really think of it as [slice,] row, column (logical)... you don't actually have to be concerned about the layout unless you want higher performance - in which case for a better access pattern you process a fundamental image-line at a time. I also find it helps me avoid bugs with xyz semantics by working with rows and columns only and remembering x=col, y = row. - I'm most familiar with having slice first like the above. - ITK is stored as row-major actually, but it's index type has dimensions specified as column,row, slice . Matlab does alot of things column order and thus acts different from implementations which can result in different outputs, but matlab seems perfectly happy living on an island where it's the only implementation providing a specific answer given a specific input. - Numpy is 0 based...? Good luck keeping it all sane though, -Jason On Sun, Aug 2, 2015 at 7:16 PM, Kang Wang kwan...@wisc.edu wrote: Thank you all for replying and providing useful insights and suggestions. The reasons I really want to use column-major are: - I am image-oriented user (not matrix-oriented, as explained in http://docs.scipy.org/doc/numpy/reference/internals.html#multidimensional-array-indexing-order-issues ) - I am so used to read/write I(x, y, z) in textbook and code, and it is very likely that if the environment (row-major environment) forces me to write I(z, y, x), I will write a bug if I am not 100% focused. When this happens, it is difficult to debug, because everything compile and build fine. You will see run time error. Depending on environment, you may get useful error message (i.e. index out of range), but sometimes you just get bad image results. - It actually has not too much to do with the actual data layout in memory. In imaging processing, especially medical imaging where I am working in, if you have a 3D image, everyone will agree that in memory, the X index is the fasted changing index, and the Z dimension (we often call it the slice dimension) has the largest stride in memory. So, if data layout is like this in memory, and image-oriented users are so used to read/write I(x,y,z), the only storage order that makes sense is column-major - I also write code in MATLAB and C/C++. In MATLAB, matrix is column-major array. In C/C++, we often use ITK, which is also column-major ( http://www.itk.org/Doxygen/html/classitk_1_1Image.html). I really prefer always read/write column-major code to minimize coding bugs related to storage order. - I also prefer index to be 0-based; however, there is nothing I can do about it for MATLAB (which is 1-based). I can see that my original thought about modifying NumPy source and re-compile is probably a bad idea. The suggestions about using fortran_zeros = partial(np.zeros(order='F')) is probably the best way so far, in my opinion, and I am going to give it a try. Again, thank you all for replying. Kang On 08/02/15, *Nathaniel Smith * n...@pobox.com wrote: On Aug 2, 2015 7:30 AM, Sturla Molden sturla.mol...@gmail.com wrote: On 02/08/15 15:55, Kang Wang wrote: Can anyone provide any insight/help? There is no default order. There was before, but now all operators control the order of their return arrays from the order of their input array. This is... overoptimistic. I would not rely on this in code that I wrote. It's true that many numpy operations do preserve the input order. But there are also many operations that don't. And which are which often changes between releases. (Not on purpose usually, but it's an easy bug to introduce. And sometimes it is done intentionally, e.g. to make functions faster. It sucks to have to make a function slower for everyone because someone somewhere is depending on memory layout default details.) And there are operations where it's not even clear what preserving order means (indexing a C array with a Fortran array, add(C, fortran), ...), and even lots of operations that intrinsically break contiguity/ordering (transpose, diagonal, slicing, swapaxes, ...), so you will end up with mixed orderings one way or another in any non-trivial program. Instead, it's better to explicitly specify order= just at the places where you care. That way your code is *locally* correct (you can check it will work by just reading the one function). The alternative is to try and enforce a *global* property on your program (everyone everywhere is very careful to only use contiguity-preserving operations, where everyone includes third party libraries like numpy and others). In software design, local invariants invariants are always better than
Re: [Numpy-discussion] Change default order to Fortran order
Thank you all for replying and providing useful insights and suggestions. The reasons I really want to use column-major are: I am image-oriented user (not matrix-oriented, as explained in http://docs.scipy.org/doc/numpy/reference/internals.html#multidimensional-array-indexing-order-issues) I am so used to read/write I(x, y, z) in textbook and code, and it is very likely that if the environment (row-major environment) forces me to write I(z, y, x), I will write a bug if I am not 100% focused. When this happens, it is difficult to debug, because everything compile and build fine. You will see run time error. Depending on environment, you may get useful error message (i.e. index out of range), but sometimes you just get bad image results. It actually has not too much to do with the actual data layout in memory. In imaging processing, especially medical imaging where I am working in, if you have a 3D image, everyone will agree that in memory, the X index is the fasted changing index, and the Z dimension (we often call it the slice dimension) has the largest stride in memory. So, if data layout is like this in memory, and image-oriented users are so used to read/write I(x,y,z), the only storage order that makes sense is column-major I also write code in MATLAB and C/C++. In MATLAB, matrix is column-major array. In C/C++, we often use ITK, which is also column-major (http://www.itk.org/Doxygen/html/classitk_1_1Image.html). I really prefer always read/write column-major code to minimize coding bugs related to storage order. I also prefer index to be 0-based; however, there is nothing I can do about it for MATLAB (which is 1-based). I can see that my original thought about modifying NumPy source and re-compile is probably a bad idea. The suggestions about using fortran_zeros = partial(np.zeros(order='F')) is probably the best way so far, in my opinion, and I am going to give it a try. Again, thank you all for replying. Kang On 08/02/15, Nathaniel Smith n...@pobox.com wrote: On Aug 2, 2015 7:30 AM, Sturla Molden sturla.mol...@gmail.com wrote: On 02/08/15 15:55, Kang Wang wrote: Can anyone provide any insight/help? There is no default order. There was before, but now all operators control the order of their return arrays from the order of their input array. This is... overoptimistic. I would not rely on this in code that I wrote. It's true that many numpy operations do preserve the input order. But there are also many operations that don't. And which are which often changes between releases. (Not on purpose usually, but it's an easy bug to introduce. And sometimes it is done intentionally, e.g. to make functions faster. It sucks to have to make a function slower for everyone because someone somewhere is depending on memory layout default details.) And there are operations where it's not even clear what preserving order means (indexing a C array with a Fortran array, add(C, fortran), ...), and even lots of operations that intrinsically break contiguity/ordering (transpose, diagonal, slicing, swapaxes, ...), so you will end up with mixed orderings one way or another in any non-trivial program. Instead, it's better to explicitly specify order= just at the places where you care. That way your code is *locally* correct (you can check it will work by just reading the one function). The alternative is to try and enforce a *global* property on your program (everyone everywhere is very careful to only use contiguity-preserving operations, where everyone includes third party libraries like numpy and others). In software design, local invariants invariants are always better than global invariants -- the most well known example is local variables versus global variables, but the principle is much broader. -n -- Kang Wang, Ph.D. Highland Ave., Room 1113 Madison, WI 53705-2275 ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
