plotting simple boiler plate plots for automatic outputs.

[samueldmcdermott]

@bnord -- right now there are functions that return arrays, which are easy to plot. I fear that making this too built in makes it less transparent (and much heavier, since we'd need to make matplotlib part of the dependency chain). Can you say more about what you want and how you'd want it to work?

[bnord]

I'd like to see some very basic diagnostic plots, like the ones that you've already made.

[samueldmcdermott]

I'm happy to save anything you think is illustrative, but can you say more about what specifically you'd like to see (plots of C_ells? flatmaps?) and how you'd like them to be stored (specific path, including a new folder if needed)?

[cavestruz]

@bnord - can you list plots in the a comment below that you'd like to see that @samueldmcdermott can check off that are not already in the example notebook? And put this into a list format? Otherwise, it looked like the example notebook has the main primary illustrative plots done and so we can close this.

[bnord]

1. Is Matplotlib a heavy dependency?

[bnord]

suggested plot list

• [ ] Cl's with different levels of noise and one cosmology
• [ ] Cl's with different cosmologies and one noise
• [ ] ibid. with maps
• [ ] ibid with maps

[samueldmcdermott]

changed the numbered list to a checklist

not sure what "heavy dependency" means. Are you asking "is matplotlib heavy?" (the answer is yes) or are you asking "is it critically important to do the stuff that our package does?" (the answer is no, it's only a small part of clplotting.py which could be deleted with no ramifications)

[cavestruz]

Just to follow up on Sam's note, having changed the numbered list to a checklist. With a checklist and the new linked branch to address the issue, GitHub provides an indicator of how close one is to being able to issue a PR (e.g. 3 out of 4 items completed). This helps with tracking feasibility and priorities if one, say, has 20min and wants to pick something out to check off.

On Fri, May 26, 2023, 12:55 samueldmcdermott @.***> wrote:

changed the numbered list to a checklist

not sure what "heavy dependency" means. Are you asking "is matplotlib heavy?" (the answer is yes) or are you asking "is it critically important to do the stuff that our package does?" (the answer is no, it's only a small part of clplotting.py which could be deleted with no ramifications)

— Reply to this email directly, view it on GitHub https://github.com/deepskies/simcmb/issues/56#issuecomment-1564666727, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABEVG65W3LP77GD4UR74EY3XIDOAHANCNFSM6AAAAAAXNAHR64 . You are receiving this because you commented.Message ID: @.***>

[bnord]

thanks, @cavestruz . I didn't know that about lists.

[bnord]

@samueldmcdermott what are best practices for including plotting libraries in science-related repositories?

[samueldmcdermott]

I'm really not sure -- because matplotlib is so heavy, and the functionality we use it for is so simple, I would vote to remove that function. But I'm not familiar with the best practice here. I'll defer to @humnaawan or @cavestruz on what to do

[bnord]

@voetberg @AeRabelais Could you chime in on this discussion? From perspective, how do you handle visualizations in code that you ship?

[voetberg]

@bnord

I think it's a fair assumption that if the end user wants to make visuals, they can make those visuals themselves and decide the specifics of the visuals.

If you're dead set on including visualization, you can always include the html file of an example notebook in documentation (reference most matplotlib example pages that include a coding example in a html)

[bnord]

I'm surprised about this, and I think I vehemently disagree, lol. But, I don't think I have enough experience to make a good argument.

@samueldmcdermott if you want to abandon or not go so far in making the viz, I'm fine with that. If I can figure out a good argument, I'll raise this issue again with more details.

[samueldmcdermott]

ok -- I'll delete the function power_spectrum_plot from clplotting.py, which is the only function that required matplotlib