First implementation of real-valued linspace.

[ejovo13]

Hi all,

This is my first time contributing on any project, ever, so bear with me if I make any missteps. As per issue #17, there seems to be a demand to implement some simple functions that return a range of values.

So, I implemented linspace using MATLAB's syntax as a guide. Currently there is no support for passing complex values into the function but if this pull request gets approved I will implement it ASAP. This pull request is a gentle attempt to make my first contribution and to understand the ropes before I start trying to tackle other open issues.

I didn't know how to implement a robust test for checking all of the elements in the returned rank 1 array, so I instead check the first and last values, and the length of the returned array in the test_linspace.f90 test.

Nevertheless, here is a quick demo when using the linspace function from another project which imports the fortran_stdlib:

[jvdp1]

Thank you and welcome!
A general comment: I wonder if such a procedure would find a better place in the module stdlib_linalg.
In Matlab it is in the same section as eye, diag,....

[milancurcic]

Thank you @ejovo13!

I wonder if such a procedure would find a better place in the module stdlib_linalg.

There's also stdlib_math which I think is more appropriate.

[ejovo13]

As per @milancurcic 's advice, I refactored the linspace interface and implementation under the stdlib_math module.
As per @jvdp1 and @arjenmarkus 's guidance, I removed the allocatable attribute from the result array and broke the function up into two different signatures, one where you pass the length of the array and the other where linspace will use the module's private paramater DEFAULT_LINSPACE_LENGTH, currently set to 100

[ejovo13]

Hello. I've implemented the functions logspace and lnspace, defined in submodules stdlib_math_logspace and stdlib_math_lnspace.

logspace

The logspace function mimicks the matlab implementation. The first two parameters are used to build the range from base^start to base^end. When base is not specified, use base 10. When the length, n, is not specified, output a rank 1 array of 50 elements.

The various function signatures of logspace (and linspace) are staggered and not using optional parameters for two reasons. The first is that using the length n as an optional parameter forces us to use allocatable arrays as our result, which should be avoided. Secondly, when using the base as an optional argument, the generic procedure is unable to distinguish function calls when the base is not passed.

this logspace implementation differs from MATLAB's because we can choose the base with which to generate the logarithmic space.

lnspace

lnspace is a special case of logspace when the base is equal to Euler's number e. Various levels of precision of EULERS_NUMBER are used in the stdlib_math module.

Tests

Running make test will check certain qualities about the results and will output a few select log files showing the arrays generated from their calls for manual inspection. Here is a gist that shows sample output.

[arjenmarkus]

How about a similar function that returns points based on a geometric series? Note: that would require an extra parameter - the ratio between adjacent intervals. Op di 1 jun. 2021 om 00:32 schreef Evan Voyles ***@***.***>:

…

Hello. I've implemented the functions logspace and lnspace, defined in submodules stdlib_math_logspace and stdlib_math_lnspace. logspace The logspace function mimicks the matlab implementation. The first two parameters are used to build the range from base^start to base^end. When base is not specified, use base 10. When the length, n, is not specified, output a rank 1 array of 50 elements. The various function signatures of logspace (and linspace) are staggered and not using optional parameters for two reasons. The first is that using the length n as an optional parameter forces us to use allocatable arrays as our result, which should be avoided. Secondly, when using the base as an optional argument, the generic procedure is unable to distinguish function calls when the base is not passed. this logspace implementation differs from MATLAB's because we can choose the base with which to generate the logarithmic space. lnspace lnspace is a special case of logspace when the base is equal to Euler's number e. Various levels of precision of EULERS_NUMBER are used in the stdlib_math module. Tests Running make test will check certain qualities about the results and will output a few select log files showing the arrays generated from their calls for manual inspection. Here is a gist <https://gist.github.com/ejovo13/c0ea7b4eefe97d1ddf0c485e8048cc20> that shows sample output. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#420 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAN6YR7UMJI3SBHKBWWUF5LTQQFATANCNFSM452VE52A> .

Requested changes addressed

[ejovo13]

Well, I believe that I've addressed, on my side, all of your comments. What I thought was going to be a quick and easy first PR has dragged on (not that that's a bad thing) and generated a lot of fruitful discussion. Thank you to everyone for their contributions and suggestions to improve this PR.

Apart from a few unresolved discussions, I don't believe that there is anything left for me to do. The ball is in your court and I will wait for any further direction. Thank you all.

[jvdp1]

LGTM.
I approve it, pending a few minor comments.
Regarding the checks, I would prefer checks on the absolute values, instead of on relative values. But I leave that for the next reviewer.

[awvwgk]

Pending any unresolved requests, this looks ready to be part of stdlib. Thanks a lot for sharing this contribution.

[jvdp1]

@ivan-pi @milancurcic you commented this PR. Could you check if your comments were answered appropriately, and appove this PR if it is ok, please?

[jvdp1]

Some details in the docs

[milancurcic]

Thank you, great addition. I only left one nit-pick about labeled format statements in test subroutines.