aclustermap 0.1.7

Convert YAMLized pandas DataFrame to Seaborn clustermap

The aclustermap package takes a YAMLized pandas DataFrame as input, with optional formatting keywords, and outputs a Seaborn clustermap .png image:

cat example.yaml | aclustermap > example.png

If YAMLizing a pandas DataFrame is a new concept, here is a quick tutorial.

Let's create a Python module generate_example.py

import pandas, yaml;

# Generate random data with make_blobs to yield interesting clustering behavior
from sklearn.datasets import make_blobs; 
example = pandas.DataFrame(make_blobs(n_samples=36, centers=3, n_features=10)[0])

# Convert the DataFrame to a dict
df_dict = example.to_dict()

# Convert the dataframe to a dict, then YAMLize the dict
yml = yaml.dump(df_dict)

# Print to stdout
print(yml)

Here is a one-liner that combines the previous code to generate this dataframe, safe it to a file, and then output it as a clustermap to example.png:

python generate_example.py > example.yaml
cat example.yaml | aclustermap > example.png

Saving the dataframe to a file is not necessary, as it can be piped directly to aclustermap:

python generate_example.py | aclustermap > example.png

aclustermap uses the simplest possible way to tweak the visualization, which is to pass a second YAML dict containing keyword argumnets for seaborn.clustermap and see also, seaborn.set_context, pyplot.setp with the Artist being the clustermap's xticklabels or yticklabels, or pyplot.rcParams.

The formatting must be enclosed in a dict labeled according to the function it applies to. Here is the default formatting, which will be used if no formatting is specified. Otherwise, simply cat a YAMLized nested dictionary with a structure similar to the following.

format:
  seaborn:
    clustermap:
      cmap: CET_CBL1
      annot: true
    set_context:
      context: notebook
  pyplot:
    setp:
      xticklabels:
        rotation: 0
      yticklabels:
        rotation: 0
    rcParams:
      font.size: 12

A convenient way to specify minor formatting tweaks directly at the command line is with a HereDoc:

(python generate_example.py; cat <<EOF) | python -m aclustermap > example.png
format:
  seaborn:
    clustermap:
      annot: false

EOF