bug for sequential transfer of same variable

[mpollow]

The following lines are copied from a notebook console. Exporting fails if the same filename is used for differently shaped data. There is no error message, but still the old data is in memory in python, see the example below. The HDF5 data on file gets updated correctly, so the bug must occur when importing to python.

By the way, what also looks like a bug is actually fine: The dimension in Matlab in the example is (4,3) and in Python (3,4). As Matlab stores the data column wise in memory and Python does so row wise, the data does not have to be resorted and is thus probably faster.

Here the example:

import pymatbridge as pymat
ip = get_ipython()
pymat.load_ipython_extension(ip)

%%matlab -o c
c = rand(4,3);

%%matlab -o c
c = rand(4,4);

c.shape
(3, 4)

[arokem]

Considering that this might be a communication issue, let's revisit this
after merging the new ZMQ communication stuff.

On Mon, Jan 20, 2014 at 2:22 PM, Chilichiller notifications@github.comwrote:

The following lines are copied from a notebook console. Exporting fails if
the same filename is used for differently shaped data. There is no error
message, but still the old data is in memory in python, see the example
below. The HDF5 data on file gets updated correctly, so the bug must occur
when importing to python.

By the way, what also looks like a bug is actually fine: The dimension in
Matlab in the example is (4,3) and in Python (3,4). As Matlab stores the
data column wise in memory and Python does so row wise, the data does not
have to be resorted and is thus probably faster.

Here the example:

import pymatbridge as pymat
ip = get_ipython()
pymat.load_ipython_extension(ip)

%%matlab -o c
c = rand(4,3);

%%matlab -o c
c = rand(4,4);

c.shape
(3, 4)

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/40
.

[mpollow]

Sounds good to me. Let me know when the new version is merged.
Thanks for your effort.

Am 22.01.2014 um 23:51 schrieb Ariel Rokem notifications@github.com:

Considering that this might be a communication issue, let's revisit this
after merging the new ZMQ communication stuff.

On Mon, Jan 20, 2014 at 2:22 PM, Chilichiller notifications@github.comwrote:

The following lines are copied from a notebook console. Exporting fails if
the same filename is used for differently shaped data. There is no error
message, but still the old data is in memory in python, see the example
below. The HDF5 data on file gets updated correctly, so the bug must occur
when importing to python.

By the way, what also looks like a bug is actually fine: The dimension in
Matlab in the example is (4,3) and in Python (3,4). As Matlab stores the
data column wise in memory and Python does so row wise, the data does not
have to be resorted and is thus probably faster.

Here the example:

import pymatbridge as pymat
ip = get_ipython()
pymat.load_ipython_extension(ip)

%%matlab -o c
c = rand(4,3);

%%matlab -o c
c = rand(4,4);

c.shape
(3, 4)

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/40
.

—
Reply to this email directly or view it on GitHub.

[arokem]

OK - sorry for how long this has taken. This seems resolved now