Pandas Dataframe Multicolor Line plot

Question

I have a Pandas Dataframe with a DateTime index and two column representing Wind Speed and ambient Temperature. Here is the data for half a day ``` temp winds 2014-06-01 00:00:00 8.754545 0.263636 2014-06-01 01:00:00 8.025000 0.291667 2014-06-01 02:00:00 7.375000 0.391667 2014-06-01 03:00:00 6.850000 0.308333 2014-06-01 04:00:00 7.150000 0.258333 2014-06-01 05:00:00 7.708333 0.375000 2014-06-01 06:00:00 9.008333 0.391667 2014-06-01 07:00:00 10.858333 0.300000 2014-06-01 08:00:00 12.616667 0.341667 2014-06-01 09:00:00 15.008333 0.308333 2014-06-01 10:00:00 17.991667 0.491667 2014-06-01 11:00:00 21.108333 0.491667 2014-06-01 12:00:00 21.866667 0.395238 ``` I would like to plot this data as one line where the color changes according to temperature. So from light red to dark red the higher the temperature for example. I found this example of multicolored lines with matplotlib but I have no idea how to use this with a pandas DataFrame. Has anyone an idea what I could do? If it is possible…

Accepted Answer

The build-in `plot` method in `pandas` probably won't be able to do it. You need to extract the data and plot them using `matplotlib`.

```
from matplotlib.collections import LineCollection
import matplotlib.dates as mpd

x=mpd.date2num(df.index.to_pydatetime())
y=df.winds.values
c=df['temp'].values
points = np.array([x, y]).T.reshape(-1, 1, 2)
segments = np.concatenate([points[:-1], points[1:]], axis=1)
lc = LineCollection(segments, cmap=plt.get_cmap('copper'), norm=plt.Normalize(0, 10))
lc.set_array(c)
lc.set_linewidth(3)
ax=plt.gca()
ax.add_collection(lc)
plt.xlim(min(x), max(x))
ax.xaxis.set_major_locator(mpd.HourLocator())
ax.xaxis.set_major_formatter(mpd.DateFormatter('%Y-%m-%d:%H:%M:%S'))
_=plt.setp(ax.xaxis.get_majorticklabels(), rotation=70 )
plt.savefig('temp.png')
```

There are two issues worth mentioning,

- the range of the color gradient is controlled by `norm=plt.Normalize(0, 10)`

- `pandas` and `matplotlib` plot time series differently, which requires the `df.index` to be converted to `float` before plotting. And by modifying the `major_locators`, we will get the `xaxis majorticklabels` back into date-time format.

The second issue may cause problem when we want to plot more than just one lines (the data will be plotted in two separate x ranges):

```
#follow what is already plotted:
df['another']=np.random.random(13)
print ax.get_xticks()
df.another.plot(ax=ax, secondary_y=True)
print ax.get_xticks(minor=True)

[ 735385.          735385.04166667  735385.08333333  735385.125
  735385.16666667  735385.20833333  735385.25        735385.29166667
  735385.33333333  735385.375       735385.41666667  735385.45833333
  735385.5       ]
[389328 389330 389332 389334 389336 389338 389340]
```

Therefore we need to do it without `.plot()` method of `pandas`:

```
ax.twinx().plot(x, df.another)
```

Pandas Dataframe Multicolor Line plot

Solution

Problem