Deprecate and remove pyplot.plotfile?

[dstansby]

Currently the record array methods in the mlab module are deprecated, but pyplot.plotfile() still requires them and has not been deprecated. Does anyone have any objections to deprecating pyplot.plotfile(), and supplying users with alternative options for reading in files (e.g. numpy, pandas)?

[ImportanceOfBeingErnest]

I guess this makes sense. The key would be really to provide alternatives. I think people like the simplicity of reading the data and plotting it in one line seen from often used constructs like

 pandas.read_csv("fname").plot()

I've never seen someone using pyplot.plotfile though. I don't know where in the docs that would fit, but showing matplotlib options like

plt.plot(*np.loadtxt("fname", unpack=True, usecols=(1,2)))

or

plt.plot("x", "y", data=pandas.read_csv("fname"))

might be helpful.

[WeatherGod]

The record array methods in mlab predates the numpy days. Why not just update plotfile() to use the numpy functions?

…

On Thu, Sep 20, 2018 at 8:14 AM Elan Ernest ***@***.***> wrote: I guess this makes sense. The key would be really to provide alternatives. I think people like the simplicity of reading the data and plotting it in one line seen from often used constructs like pandas.read_csv("fname").plot() I've never seen someone using pyplot.plotfile though. I don't know where in the docs that would fit, but showing matplotlib options like plt.plot(*np.loadtxt("fname", unpack=True, usecols=(1,2))) or plt.plot("x", "y", data=pandas.read_csv("fname")) might be helpful. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#12180 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AARy-ET2e4DjXTlHtsTs_BY8eO0vg2LCks5uc4aJgaJpZM4Wx5pR> .

[timhoffm]

+0.5 on deprecating plotfile(). It does too much. Parsing a file and plotting are separate concerns. While it may be convenient for a user, it's not reasonable from an API point of view.

pandas.read_csv("fname").plot() does it right by chaining the different functionalities. The other two pyplot-based code snippets by @ImportanceOfBeingErnest are reasonable as well. We could just add an example for "Plotting data from a file" to promote the alternatives.

The focus of matplotlib is plotting, not parsing data out of a file. The latter is a complicated topic itself and we should not bother too much about it ourselves.

That said, I don't mind if someone has strong feelings about plotfile() and wants to take on the burden to port it to numpy.

[dstansby]

For now I have just made csv2rec a private method (well, added an underscore to its name), which is then used in plotfile(). If anyone else wants to replace it then feel free to open a new PR.

[rudrathegreat]

Another alternative to matplotlib.pyplot.plotfile is matplotlib.pyplot.savefig which allows you to save a figure (graph) into a folder. Simply -

plt.savefig('Images/image.png')

Requirements

Python 3.7

Matplotlib 2.2.3

SciPy (if you want to save the data as an audio file)

NumPy (recommendation, if you want speed)

I hope that helped!

[WeatherGod]

@rudrathegreat, uhm. I don't follow. savefig() is not an alternative to plotfile(). And I don't know why you included a requirements list there.

[jklymak]

@rudrathegreat Thanks so much for your interest in Matplotlib.

When calibrating the kind of comments to post here, please note whether commenters are on the development team or not, as indicated by a "Member" badge in the heading of their posts. Most folks who are on the development team are aware of something as basic as fig.savefig.

[rudrathegreat]

@jklymak and @WeatherGod . Thanks for letting me know. I thought the issue was related to saving the graph. I love Matplotlib! Thanks for letting me know!

[anntzer]

Done in #13740.