ENH: adds new Function.savetxt method

[Gui-FernandesBR]

Pull request type

• Code changes (bugfix, features)

Checklist

• Tests for the changes have been added (if needed)
• Docs have been reviewed and added / updated
• Lint (black rocketpy/ tests/) has passed locally
• All tests (pytest --runslow) have passed locally
• CHANGELOG.md has been updated (if relevant)

Current behavior

If you have a Function object, good luck trying to export this to be used later or to send to someone else.

New behavior

Inspired by the numpy.savetxt method, I've created the Function.savetxt that is currently working very well for both 1D or N-D Function, with or without discrete sources.

Breaking change

• No

Additional information

• In the future we may need a to_serializable in other to export different attributes of the Function instances to a .json file. Let's leave this for a future PR.
• The tests are running really fast... thank you numpy for exsisting.
• Also, can I say how better the Function classes are now after we officialized the support for csv files with headers? Amazing feature!

[codecov]

Codecov Report

Attention: 2 lines in your changes are missing coverage. Please review.

Comparison is base (2205fdf) 72.20% compared to head (696e709) 72.23%.
Report is 1 commits behind head on develop.

Files	Patch %	Lines
rocketpy/mathutils/function.py	85.71%	2 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #514      +/-   ##
===========================================
+ Coverage    72.20%   72.23%   +0.02%     
===========================================
  Files           56       56              
  Lines         9376     9389      +13     
===========================================
+ Hits          6770     6782      +12     
- Misses        2606     2607       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[phmbressan]

Really good PR, complete and with a nice set of tests. There is one detail that I believe is important to be looked upon: the set_discrete changes the source.

[MateusStano]

What do you think of having to_txt and to_csv methods instead of just savetxt?
This would make sense if we were ever to add a way to save a function to a .json for example, with to_json

It seems you got savetxt from numpy. The to_ functions come from pandas

This is just a loose suggestion. I don't mind the current name

[MateusStano]

Actually, just saw what you commented in the description. I still think it could be more descriptive to use to_ names

[Gui-FernandesBR]

This is an interesting suggestion.

Between numpy and pandas, I have to say that numpy might be a better inspiration in this case, since pandas is not one of the rocketpy's dependency.

Beyond that, I believe we need to be careful with the to_ methods, let's see:

• .savetxt: takes a Function instance and save its source to a text file, where each row represents one point from the source.
• .savecsv: this call the savetxt method above but modify the file extension to .csv
• .to_json (or other similar name like ``) takes the Function instance and create a json file with different attributes from the object (e.g. interpolation method, inputs and outputs names, repr method, etc.)

If we change the savetxt to to_txt we would be creating an opportunity for confusion since the to_txt and to_json would export totally different contents.
There's a reason for having the save methods to save the source and the to_ methods to convert the entire object to something else.

Let me know in case your disagree with this.

The #522 might be connected to this comment too.

[Gui-FernandesBR]

All fixed here.
@MateusStano @phmbressan could you please re-review it?
Thank you for the past comments btw, I'm happy that you brought my attention to certain points.

[phmbressan]

As I mentioned in my last review, really good, this PR makes it really easy to export data in RocketPy. Nice to see that #519 was indeed useful.