victorlei / smop Goto Github PK

Matlab operator * is the dot product when multiplying matrices or vectors
and is translated to python operator * which is the element-wise multiplication when
using matrices or vectors; while matlab operator .* which is the element-wise multiplication is translated to .dot()

Comments annotate code as a docstring and at the few places where the code itself does not suffice for readability. For this reason it would be nice to provide an option for converting MATLAB comments to python comments. This was mentioned also in #12 and in this issue on chiselapp.

Matlab's operator = deep copies the matrix, while in numpy it's only the reference that is copied, even when affecting a sub part of the matrix/vector.
Fix:
a = b
should be:
a = np.array(b)
when b is a vector/matrix.

But it's not useful when affecting the result of an operation:
a = b * b
for example.

PS: adding spaces right and left of the = operator would make code more readable, as keeping comments and line feeds.

perhaps adding 'convert' to the project description would increase the probability of google suggesting it for a "convert matlab to python" search

lenght function should be converted to len

currently:

arr{1} = 1;
arr{2} = 2;
lentgh(arr);

converted to

arr[1] = 1
arr[1] = 1
lenght(arr)

There are several other projects that translate matlab to python. The most relevant seems to be these two, also written in python and BSD / MIT licensed:
https://bitbucket.org/juricap/ompc/src
https://github.com/miaoever/Mat2py

I'm not certain which project is furthest along (I chose this one to bug since it was the most recent), but working on finishing one codebase to perfection rather than having three different but incomplete ones ought to be much more efficient...

function foo(x)
foo = 10
end

Example:

function a = test(b)
         [c,d] = butter(4, 'low')
         a = filter(c,d, b)
         a = round(a)
end

gets converted to:

def test(b):
    c, d = butter(4, 'low') # nargout=2
    a = filter_(c, d, b)
    a = round_(a)
    return a

This happens both in the master branch and @perimosocordiae's divergent branch.

Any pointers?

Maybe even in the home page would be very helpful. Though i am sure it would be done pretty soon, but having some demo will generate more interest!

Since ply is developed separately, importing it from site-packages will benefit from any updates and upgrades it experiences.

The docs recommend to use argparse instead of getopt.

Running Red Hat Linux:

$ smop

Does not display the text in usage() of main.py.

This Matlab function:

    function a = test(m)
             a = m.member(:, 0)
    end

produces this python function:

    def test(m):
        a = m.member(:, 0)
        return a

The obvious conclusion I can make is that smop is detecting any symbol with '.' in it as a function. Any idea what would need to be changed to fix this?

• cython?
• arrays with [ ], structs
• command mode
• comments
• dot op
• fastsplver
• compare performance --> profiler
• .py3 compatibility
• pep8
• ompc examples
• libermate examples
• matlab2fortran -> regex.m low prio
• pypy

matlab:

function ret = noargin()
    ret = 0

python:

# Autogenerated with SMOP version 0.25
# main.py test.m -o test.py
from __future__ import division
import numpy as np
from runtime import *
def noargin_(,nargout=1):
    #3
    ret=0
    return ret

For unknown reason, UNC path combined with asterisk, as in smop \foo\bar\bzz*.m is not understood and somehow prevents smop from being run. Other, similar, expressions work just fine. For example, smop "\foo\bar\bzz*.m" (note the quotes) is okay.

Hi all,
I am new to Python and need to convert some MATLAB code to Python code, so I tried the "Working example: solver.m" on https://github.com/victorlei/smop/blob/master/README.rst as a start, but got the error message:

~/Downloads/smop-0.25.4/smop> python main.py solver.m
Traceback (most recent call last):
File "main.py", line 164, in
main()
File "main.py", line 157, in main
G = resolve.resolve(func_obj)
File "/nfs/yugong/data/jnwang/Downloads/smop-0.25.4/smop/resolve.py", line 65, in resolve
G = as_networkx(t)
File "/nfs/yugong/data/jnwang/Downloads/smop-0.25.4/smop/resolve.py", line 50, in as_networkx
G.add_node(uu, ident=u) # label=label)
TypeError: add_node() got an unexpected keyword argument 'ident'

I also tried the latest version smop-0.25.6, and ended up with the same error, anyone knows what is the problem? I am using openSUSE 13.1, python 2.7.6, and MATLAB R2012b.

Thanks a lot!

The generated code needs spaces and other adjustments to become PEP8 conformant, hence more readable. Issue opened following comment in #12.

=>> function x=foo(y); x=y; end
Traceback (most recent call last):
File "main.py", line 139, in
main()
File "main.py", line 93, in main
resolve.resolve(t,symtab)
File "/home/lei/smop-github/smop/resolve.py", line 67, in resolve
u = G.node[n]["ident"]
KeyError: 'ident'
lei@dilbert ~/smop-github/smop $

It would be nice to make smop compatible with python3, since ply is.

I accidentally uploaded a development version -- sorry I have no experience with git... Meanwhile download version 0.22 -- it is supposed to work . In particular, you should be able to run

python -m fastsolver.py

Without error messages.

what simple and obvious step did I miss?

If you ever tried to understand the output from the parser, you will appreciate -d option. Take a look at solver.pdf for an example.