Blog - bar

_posts/altair-to-mpl-bar.md

@@ -0,0 +1,90 @@
+---
+layout: post
+title:  "Making a Bar Chart"
+date:   2018-08-15 16:00:00 -0500
+author: "Kimberly Orr and Nabarun Pal"
+categories: user-guide
+tags: "intro about bar"
+excerpt_separator: <!--read more-->
+---
+
+# Making a Bar Chart
+At the time of writing, mpl-altair does not support bar charts, so this post will show how to create a bar chart in Altair, Matplotlib, and how mpl-altair _should_ implement bar chart conversion in the future.
+
+We'll work with the following long-form DataFrame for this example:
+```python
+import pandas as pd
+df = pd.DataFrame({
+    'group': ['1', '1', '2', '2', '3', '3', '4', '4', '5', '5'],
+    'variable': ['a', 'b', 'a', 'b', 'a', 'b', 'a', 'b', 'a', 'b'],
+    'scores': [20, 25, 35, 32, 30, 34, 35, 20, 27, 25]
+})
+```
+
+## Altair
+For this dataset, specifying the color will automatically stack the bar charts.
+```python
+import altair as alt
+alt.Chart(df).mark_bar().encode(
+    x='group',
+    y='scores',
+    color='variable'
+)
+```
+![png](pics/altair-to-mpl-bar_0.png)
+
+## Matplotlib
+This is a little more complicated in Matplotlib. Since Matplotlib is procedural, we have to manually tell Matplotlib to stack the bars. Also notice that we are calling a new function now (`ax.bar()`) to get a bar plot.
+
+To stack the bars, we have to create new subset dataframes and then plot each one separately (specifying that bottom of one plot should be the top of another `bottom=a_scores`). One way to do this is to subset via indexing (option 1). Another way to do this is to use the `df.groupby()` function (option 2).
+```python
+import matplotlib.pyplot as plt
+```
+```python
+# Option 1
+fig, ax = plt.subplots()
+groups = df['group'].unique()
+a_scores = df[df['variable']=='a']['scores']
+b_scores = df[df['variable']=='b']['scores']
+ax.bar(groups, a_scores, label='a')
+ax.bar(groups, b_scores, bottom=a_scores, label='b')
+ax.set_xlabel('groups')
+ax.set_ylabel('scores')
+ax.legend()
+plt.grid()
+plt.show()
+```
+```python
+# Option 2
+fig, ax = plt.subplots()
+
+(_, a), (_, b) = df.groupby('variable')
+ax.bar(a['group'], a['scores'], label='a')
+ax.bar(b['group'], b['scores'], bottom=a['scores'], label='b')
+ax.set_xlabel('groups')
+ax.set_ylabel('scores')
+ax.legend()
+plt.grid()
+plt.show()
+```
+Both produce:
+
+![png](pics/altair-to-mpl-bar_1.png)
+
+
+## mpl-altair
+At the time of writing, mpl-altair doesn't support bar charts.
+
+If mpl-altair supported bar charts, this is how an Altair chart would get rendered in mpl-altair:
+```python
+import altair as alt
+import matplotlib.pyplot as plt
+import mplaltair
+chart = alt.Chart(df).mark_bar().encode(
+    x='group',
+    y='scores',
+    color='variable'
+)
+fig, ax = mplaltair.convert(chart)
+plt.show()
+```