command line tool

[chanansh]

Is anyone planning a command line tool for librosa such as the one in yaafe?
If not I will be happy to help

[bmcfee]

Not as such, but it sounds like a great idea!

[chanansh]

I have problems running the tests on the master branch. I am trying to follow the contribution guide but cannot make the nose test run. On May 26, 2017 5:57 PM, "Brian McFee" <notifications@github.com> wrote: Not as such, but it sounds like a great idea! — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#582 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AFtFzFJzzsR5ZPB4oy2oMfWx8_VRzxekks5r9uhJgaJpZM4NnT-K> .

[lostanlen]

I have given some thought to this today. The problem is that most functions in librosa return arrays. For functions that return integer, strings, or nothing, it's conceivable to print these to the command line. However, for arrays, these outputs should be exported. So we need to agree on a format for storing arrays and use it consistently.
I can see three options:
(1) numpy.save. Pros: relatively lightweight, no need to add a dependency. Cons: not interoperable.
(2) h5py. Pros: lightweight, interoperable. Cons: need to add a dependency.
(3) csv. Pros: human readable, no need to add a dependency. Cons: slow, only makes sense for 2D arrays.

[lostanlen]

I'm thinking about this issue again and i don't really see a compelling use case for it. Given how easy it is to call Python scripts from the command line, is this really needed? I feel like it would add a lot of boilerplate code on our end.

[bmcfee]

I agree with @lostanlen -- librosa is not intended to be the end-point of audio analysis, but rather the set of building blocks that you would use to make such a thing. Any command-line application that we design around the library will undoubtedly fail to meet the needs of most users, so I think it's best that we stay out of that space.