How can I efficiently extract elements from a large cell in MATLAB?

[member 428835]

Hi PF!

I have a cell 1 X 5001. Each component within is a double matrix of different sizes, approximately 100 X 3. The entire cell is called a. Then when trying to extract component (1,2) from each of the 5001 matrices I type a{:}(1,2), but upon executing this MATLAB complains "Expected one output from a curly brace or dot indexing expression, but there were 5001 results." Any ideas?

 

[jedishrfu]

Heres a Matlab reference on brace notation for data access

https://www.mathworks.com/help/matlab/matlab_prog/access-data-in-a-table.html
it seems you are doing it wrong from the reference example.

Being a java programmer, I’ve always had trouble with these slicing and dicing operations and always fall back on individual cell indexing to get my data.

 

[Wrichik Basu]

Being a java programmer, I’ve always had trouble with these slicing and dicing operations and always fall back on individual cell indexing to get my data.

I second that. I apways forget those syntaxes, whether in Matlab or in Python. Simple cell indexing using loops is always better.

 

[jedishrfu]

Although arguably slower too since matlab's syntax allows for vectorization of operations.

 

[member 428835]

Thanks for all the comments. Ended up just using a for loop, though I agree with jedishrfu statement on vectorization. Would've been nice but fortunately this computation takes about 0.1 seconds so no loss there.

 

[jedishrfu]

Yeah its good to know these quirky operators in MATLAB and other languages. They added to optimize and sometimes provide a cool feature to discourage migration to something else.

An expert MATLAB programmer will know these shortcuts and can whip out code faster and will laugh upon seeing the for loop stuff.

 

[member 428835]

Yeah its good to know these quirky operators in MATLAB and other languages. They added to optimize and sometimes provide a cool feature to discourage migration to something else.

An expert MATLAB programmer will know these shortcuts and can whip out code faster and will laugh upon seeing the for loop stuff.

Yea no kidding. I realized a while ago the people who know their stuff well make me look like a noob...I probably am anyways

 

[SpinFlop]

For future reference, when performing a uniform operation on arrays inside of a cell, just write a function to perform your desired task and have it operate across the cell using cellfun. For your current problem it would look like this:
cellfun(@(f)f(1,2), a)

 

[Dr Transport]

Yea no kidding. I realized a while ago the people who know their stuff well make me look like a noob...I probably am anyways

Don't beat yourself up. I work across two groups at work, my home group has zero skills outside of FORTRAN 90/95 and they don't even use all the bells and whistles. The older guys in the other group try to write in MATLAB, but it looks a lot like FORTRAN, loops all over the place and they wonder why their codes runs soooo slowwww. I grabbed on short script, vectorized it and realized a 50% time savings, only took me a week to do it. They asked another group to write them some better software for their data sets, he came back with OOP MATLAB, they were completely lost since they only had a minimal amount of C++.

 

[member 428835]

Don't beat yourself up. I work across two groups at work, my home group has zero skills outside of FORTRAN 90/95 and they don't even use all the bells and whistles. The older guys in the other group try to write in MATLAB, but it looks a lot like FORTRAN, loops all over the place and they wonder why their codes runs soooo slowwww. I grabbed on short script, vectorized it and realized a 50% time savings, only took me a week to do it. They asked another group to write them some better software for their data sets, he came back with OOP MATLAB, they were completely lost since they only had a minimal amount of C++.

Wow, even in the "real world" people still run into large speed bumps. This was shockingly very nice to read: thanks!