If you have a license of MATLAB, you can use the gallery function to generate some very good test matrices. *Rajaditya Mukherjee * *3rd Year Graduate Student* Dept. of Computer Science and Engineering The Ohio State University Columbus, Ohio Tel :- +1-(614)-271-4439 email :- rajaditya.mukherjee@gmail.com , raj@rajaditya.com On Tue, Mar 10, 2015 at 8:18 AM, Ganesh Prasad wrote:

Hey Dhruv I am an applicant myself, I think I can give you a few ideas regarding your queries.

...

1. Can someone please direct me to some code snippet that shows how the error handling code is written in boost library.

Go through the library source code thoroughly. Use a code dependency graph generator to browse code easily. (Easiest way is to run doxygen over the ublas directory). That will make going through the code real fast.

...

2. What sort of test code is required? Do we need to have a test case with hard-coded output to be checked against the results obtained when the test case is passed to the algo. Or there's some other predefined format for testing codes in boost library.

Not sure about the format, but the tests should cover as many cases as possible.

Best wishes Ganesh Prasad

On 10 March 2015 at 15:03, Dhruv Kohli wrote:

...

Hello,

I am Dhruv, majoring in Mathematics & Computing from Indian Institute of Technology, Guwahati. After doing a bit of research over the project ideas, I found the idea of further developing boost.ublas library interesting and have started writing the programming competency code. Here is the link to the code I've written yet:

https://github.com/chiggum/boost.ublas/blob/master/gaussian_elimination.hpp

...

. Though the code is not yet complete, but a naive version of algorithm is implemented and is integrated into ublas. To me, the code seems to be running properly.

I need a little help for proceeding further. 1. Can someone please direct me to some code snippet that shows how the error handling code is written in boost library. To be specific, the gaussian elimination doesn't work on singular matrices. So, I would like to set the errno and display appropriate message to the user whenever a singular matrix is given as an input.

2. What sort of test code is required? Do we need to have a test case with hard-coded output to be checked against the results obtained when the test case is passed to the algo. Or there's some other predefined format for testing codes in boost library. Sorry, I haven't yet gone through most of the code in boost library.

Any further helpful links/resources are welcomed.

Thanks & Best regards, Dhruv https://github.com/chiggum

P.S. Sorry for reposting the mail. I am still waiting for a reply.

_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

Thanks a lot guys for replying. Using doxygen and the information at ublas wiki, I was able to understand the error handling and test case format.

​ Hi guys, it seems I will have dozens of great students already working as a community. Google please give me a slots for each of them ! ​ ​Hi Dhruv, thanks for your contribution. I will review it quickly too. It's about time to write your proposal too. The sooner the better. Best, David​

For reporting singularities or similar conditions, there are really only two options, either throw an exception or return an error code, but only one of those two must be used consistently throughout uBlas. I personally prefer to work with exceptions, but I know that in this application domain it can be a very hard sell (especially with people wanting to use this in embedded systems). So, that really only leaves us with error-codes. That's what I would recommend.

​I think you're pointing to a complex problem. Indeed, assertions are usually ​ ​not a good idea and your use case is typical and quite common in fact. ___[For the potential GSOC students]___ Should we use exceptions or return codes ?​ ​ ​Again, as you pointed, exceptions are great, with almost no overhead except when there is an ... exception. And most of the time, the execution time ​is hard to predict. Which is far from being a good idea in a real-time system. I can't really say who the ublas users are, because there are a lot in all sort of applications. Now, think about these 2 cases: - you implement a multiplication algorithm for 2 matrices: you want your algorithm to return the result and to be embedable into another expression, that is you want to be able to write C = A + (B*C). So it's important that the return value is the result itself. Your only option to report an error is to throw an exception. But it's hard to have a problem with a multiplication (hard but not impossible), - you implement a decomposition algorithm, and as a result you want 2 matrices (for example). You know that even for simple cases, you're going to run a good and long algorithm. And have 2 values (i.e. 2 matrices) in return. So it wouldn't be a bad idea to pass 2 empty matrices by reference where to store the result (it's just an example, your design may vary), and return a code which says if things went wrong or not. It could be a boolean value, an enum, or anything that is quickly interpretable. And you really don't want to include your decomposition algorithm into a bigger expression. You can simply decompose your complex calculations and have one line dedicated to the use of this decomposition algorithm. Therefore, I think a return code is a better idea.

On 15 Mar 2015 at 16:59, Mikael Persson wrote:

...

Should we use exceptions or return codes ?

I agree with all that you've said on this. It's very true that one of the nice benefits of exceptions is that they allow you to retain the nice functional / expression syntax, that is, instead of having to do each operation on its own line, grabbing the error-code and checking it.

The lightweight non-allocating future I proposed here many months ago is functional, returns any unexpected type E you like, and the compiler should completely elide the future code if it isn't necessary. In other words, you write your code with promise-futures, and only if they cross a thread boundary does the compiler actually generate any code for them. I decided against implementing those properly, but AFIO v1.4's afio::future<T> will be non-allocating and lightweight (i.e. elidable by the optimiser). It also will carry either a type T, an error_code or an exception_ptr, so it's a fixed function constexpr variant transport. I'll be donating that afio::future<T> to Boost.Thread v5 probably as lightweight_future<T>. Someone else may decide to turn it into a totally generic solution suitable for replacing future/promise in C++ 17. Or they may not. Niall -- ned Productions Limited Consulting http://www.nedproductions.biz/ http://ie.linkedin.com/in/nialldouglas/

If you don't have one, I think that OCTAVE maybe a good alternative for MATLAB 2015-03-10 21:18 GMT+08:00 Rajaditya Mukherjee < rajaditya.mukherjee@gmail.com>:

If you have a license of MATLAB, you can use the gallery function to generate some very good test matrices.

*Rajaditya Mukherjee * *3rd Year Graduate Student* Dept. of Computer Science and Engineering The Ohio State University Columbus, Ohio Tel :- +1-(614)-271-4439 email :- rajaditya.mukherjee@gmail.com , raj@rajaditya.com

On Tue, Mar 10, 2015 at 8:18 AM, Ganesh Prasad wrote:

...

Hey Dhruv I am an applicant myself, I think I can give you a few ideas regarding your queries.

...

1. Can someone please direct me to some code snippet that shows how the error handling code is written in boost library.

Go through the library source code thoroughly. Use a code dependency graph generator to browse code easily. (Easiest way is to run doxygen over the ublas directory). That will make going through the code real fast.

...

2. What sort of test code is required? Do we need to have a test case with hard-coded output to be checked against the results obtained when the test case is passed to the algo. Or there's some other predefined format for testing codes in boost library.

Not sure about the format, but the tests should cover as many cases as possible.

Best wishes Ganesh Prasad

On 10 March 2015 at 15:03, Dhruv Kohli wrote:

...

Hello,

I am Dhruv, majoring in Mathematics & Computing from Indian Institute of Technology, Guwahati. After doing a bit of research over the project ideas, I found the idea of further developing boost.ublas library interesting and have started writing the programming competency code. Here is the link to the code I've written yet:

...

...

. Though the code is not yet complete, but a naive version of algorithm is implemented and is integrated into ublas. To me, the code seems to be running properly.

I need a little help for proceeding further. 1. Can someone please direct me to some code snippet that shows how the error handling code is written in boost library. To be specific, the gaussian elimination doesn't work on singular matrices. So, I would

https://github.com/chiggum/boost.ublas/blob/master/gaussian_elimination.hpp like

...

to

...

set the errno and display appropriate message to the user whenever a singular matrix is given as an input.

2. What sort of test code is required? Do we need to have a test case with hard-coded output to be checked against the results obtained when the test case is passed to the algo. Or there's some other predefined format for testing codes in boost library. Sorry, I haven't yet gone through most of the code in boost library.

Any further helpful links/resources are welcomed.

Thanks & Best regards, Dhruv https://github.com/chiggum

P.S. Sorry for reposting the mail. I am still waiting for a reply.

_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

_______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

-- with kind regards ------------------------------------------------------------------ Yan Ni, undergraduate of Department of Computer Science, School of Electronic Engineering & Computer Science, Peking University