Seaborn

[hadikazad-pixel]

Is there anyway, to use the Seaborn libraries in the Pythonista?
Bought it to using the data science research and analysis
Now I can’t overcome the problem of not having Seaborn libraries
How can I install these libraries into my IDE?
If not, I ask developers to add this feature into future update
For me, it’s a big problem, not having seaborn and doing the visualization of the data

[mmccthomas]

Hi,
In theory you can install seaborn as it is 100% Python, with dependencies on matplotlib, pandas and numpy. All these are in Pythonista.
However, performing a ‘pip install’ of seaborn throws up some version incompatibility between latest pandas and numpy.
I would suggest getting seaborn from GitHub via the Code/Download Zip function.
Unzip the file in Files/iCloudDrive/Downloads and drag and drop seaborn folder into Pythonista Python Modules/Site Packages(user).
This is the manual way to get new modules into Pythonista.

Note: if anyone else can advise on how to safely remove folder /pandas from site-packages(user) I’d be grateful!

Chris

[RichardPotthoff]

I would just "swipe left -> Trash" to delete files/folders from "site-packages(user)" if "pip uninstall" does not work.

[mmccthomas]

Thanks Richard.
Some time ago I emptied the Trash folder programmatically and it no longer functions. Trashed files and folders reappear after a few seconds. I fear that I will need to delete and reinstall Pythonista to fix, which fills me with tredipation. Or wait for the much anticipated Pythonista update(!). I have been using os and shutils instead.

[RichardPotthoff]

Hi Chris,
before you reinstall Pythonista, you could try to start Pythonista in safe mode:
#53 (comment)
You could also investigate if there is something wrong with pip:

pip help (for a list of useful commands)
pip config list
pip check
pip inspect

I also found "pipdeptree" useful:

pip_sh:site-packages $ pipdeptree -p pipdeptree
pipdeptree==2.28.0
  - packaging [required: >=24.1, installed: 25.0]
  - pip [required: >=24.2, installed: 25.2]
pip_sh:site-packages $ 

[mmccthomas]

Thanks Richard, that solved an issue where installing seaborn via stash had created a new pandas folder in site-packages which was not compatible with the installed numpy. I deleted this folder but a module of mine which used pandas then failed. Starting in safe mode fixed that. So thanks for the suggestion. Chris

[RichardPotthoff]

The problem is the build process that some packages require. Often it is a dependency that causes the problem, and not the main package itself (in your case "seaborn").

I had some success using the "--no-deps" option with pip install:

pip install --no-deps seaborn

Another option is to install the package on another computer, and copy the installed package folders (e.g. seaborn, seaborn-0.13.2.dist-info) to pythonista's "site-packages(user)" folder. Here is the Python script I use:

#80 (comment)

This method can also be used to install "pip" itself, if it is not already installed.

[RichardPotthoff]

I have installed seaborn on my iPad with the --no-deps option, and it turns out that the matplotlib version that comes pre-installed with Pythonista 3.4 is not quite up-to-date enough for the latest version of seaborn:

seaborn==0.13.2
  - numpy [required: >=1.20,!=1.24.0, installed: 1.22.3]
  - pandas [required: >=1.2, installed: 1.4.4]
  - matplotlib [required: >=3.4,!=3.6.1, installed: 3.3.3+0.g5a4f1b675d.dirty]

This could cause problems with some newer features of seaborn (according to Grok):

• API Incompatibilities:

  • Seaborn 0.13.2 expects Matplotlib 3.4+ features (e.g., improved PathCollection handling, better Figure defaults).
  • Functions like sns.scatterplot, sns.lineplot, or sns.histplot may raise:
    • AttributeError: 'PathCollection' object has no attribute 'set_offset_position' or
    • TypeError: Figure.add_subplot() got an unexpected keyword argument 'projection'

• Missing Features:

  • No support for tight_layout(pad=...) improvements.
  • Broken sns.catplot with kind="swarm" on older backends.

If this becomes a problem, installing an older version of seaborn could be be a solution:

pip uninstall seaborn
pip install --no-deps "seaborn<0.13"

For pipdeptree -p seaborn I then get:

seaborn==0.12.2
  - numpy [required: >=1.17,!=1.24.0, installed: 1.22.3]
  - pandas [required: >=0.25, installed: 1.4.4]
  - matplotlib [required: >=3.1,!=3.6.1, installed: 3.3.3+0.g5a4f1b675d.dirty]