PolyMath v0.93

Hi all,

I release a new version of PolyMath with some pending commits in the inbox. You can load the new version with :

Metacello new 
  repository: 'http://smalltalkhub.com/mc/PolyMath/PolyMath/main';
  configuration: 'PolyMath';
  version: '0.93';
  load

Apparently there is some red tests with matrices. Evan can you have a look, apparently this is your code ?

​Thank you all.

​

--

Serge Stinckwich
UMI UMMISCO 209 (IRD/UPMC/UY1)

"Programs must be written for people to read, and only incidentally for machines to execute."

http://www.doesnotunderstand.org/

On Sun, Dec 31, 2017 at 12:23 AM, Evan Donahue <[hidden email]> wrote:

Hi,

​Hi Evan,

​

 

No problem. It looks like I wrote the tests, but not the code. I wrote the tests to document bugs in the existing code.

​Sorry Evan, I thought it was you. My mistake.

This is really great to wrote tests about the API !

We need more of these :-)

 

- PMMatrix >> cumsum has been changed so that it returns self (collect has been replaced with do), which looks to me like it is a mistake.

​I fix this one. Apparently the code was lost during some previous merges ...

​Available now in Math-Matrix-SergeStinckwich.25

- Adding numbers and matrices no longer performs element-wise addition, but seems to throw an error. Is this intended behavior? I could see an argument for either way. I prefer the polymorphic mixing of numbers and matrices, when possible.

- Adding matrices of different sizes returns an incorrect matrix rather than throwing a SizeMismatch error. This could be intended behavior or not depending on whether you want to do array bounds checking or not for speed. I prefer the former, but I know others prefer the latter. Ideally this could be solved with the contracts library I should be releasing soon and everyone will be happy.

So as far as I can tell, the tests still look right to me unless the underlying API has changed. I don't know who wrote or changed the methods, so I'm not sure what their intended behavior is supposed to be. I can help patch them, but I'm not sure how to decide what the correct behavior is supposed to be in cases where it is unclear. I don't want to just change them back if someone had a reason to try to change them in the first place. Thoughts?

​I have to look at these one.

​

​Thank you for you work.

​

Best,

Evan

On Sat, Dec 30, 2017 at 5:48 PM, Serge Stinckwich <[hidden email]> wrote:

Hi all,

I release a new version of PolyMath with some pending commits in the inbox. You can load the new version with :

Metacello new 
  repository: 'http://smalltalkhub.com/mc/PolyMath/PolyMath/main';
  configuration: 'PolyMath';
  version: '0.93';
  load

Apparently there is some red tests with matrices. Evan can you have a look, apparently this is your code ?

​Thank you all.

​

--

Serge Stinckwich
UMI UMMISCO 209 (IRD/UPMC/UY1)

"Programs must be written for people to read, and only incidentally for machines to execute."

http://www.doesnotunderstand.org/

--

Serge Stinckwich
UMI UMMISCO 209 (IRD/UPMC/UY1)

"Programs must be written for people to read, and only incidentally for machines to execute."

http://www.doesnotunderstand.org/

Hmm. I would bet that if I wrote tests, I probably also submitted patches to make the tests pass. Might those have been rolled back in the merge as well? Unless someone has expressly changed those methods, it sounds like they may have just reverted.

Evan

On Sun, Dec 31, 2017 at 5:29 AM, Serge Stinckwich <[hidden email]> wrote:

On Sun, Dec 31, 2017 at 12:23 AM, Evan Donahue <[hidden email]> wrote:

Hi,

​Hi Evan,

​

 

No problem. It looks like I wrote the tests, but not the code. I wrote the tests to document bugs in the existing code.

​Sorry Evan, I thought it was you. My mistake.

This is really great to wrote tests about the API !

We need more of these :-)

 

- PMMatrix >> cumsum has been changed so that it returns self (collect has been replaced with do), which looks to me like it is a mistake.

​I fix this one. Apparently the code was lost during some previous merges ...

​Available now in Math-Matrix-SergeStinckwich.25

- Adding numbers and matrices no longer performs element-wise addition, but seems to throw an error. Is this intended behavior? I could see an argument for either way. I prefer the polymorphic mixing of numbers and matrices, when possible.

- Adding matrices of different sizes returns an incorrect matrix rather than throwing a SizeMismatch error. This could be intended behavior or not depending on whether you want to do array bounds checking or not for speed. I prefer the former, but I know others prefer the latter. Ideally this could be solved with the contracts library I should be releasing soon and everyone will be happy.

So as far as I can tell, the tests still look right to me unless the underlying API has changed. I don't know who wrote or changed the methods, so I'm not sure what their intended behavior is supposed to be. I can help patch them, but I'm not sure how to decide what the correct behavior is supposed to be in cases where it is unclear. I don't want to just change them back if someone had a reason to try to change them in the first place. Thoughts?

​I have to look at these one.

​

​Thank you for you work.

​

Best,

Evan

On Sat, Dec 30, 2017 at 5:48 PM, Serge Stinckwich <[hidden email]> wrote:

Hi all,

I release a new version of PolyMath with some pending commits in the inbox. You can load the new version with :

Metacello new 
  repository: 'http://smalltalkhub.com/mc/PolyMath/PolyMath/main';
  configuration: 'PolyMath';
  version: '0.93';
  load

Apparently there is some red tests with matrices. Evan can you have a look, apparently this is your code ?

​Thank you all.

​

--

Serge Stinckwich
UMI UMMISCO 209 (IRD/UPMC/UY1)

"Programs must be written for people to read, and only incidentally for machines to execute."

http://www.doesnotunderstand.org/

--

Serge Stinckwich
UMI UMMISCO 209 (IRD/UPMC/UY1)

"Programs must be written for people to read, and only incidentally for machines to execute."

http://www.doesnotunderstand.org/