python - Seaborn FacetGrid barplots and hue -

i have dataframe following structure:

interval    segment  variable        value 4   02:00:00      night  weekdays   154.866667 5   02:30:00      night  weekdays   100.666667 6   03:00:00      night  weekdays    75.400000 7   03:30:00      night  weekdays    56.533333 8   04:00:00      night  weekdays    55.000000 9   04:30:00      night  weekends    53.733333 10  05:00:00      night  weekends    81.200000 11  05:30:00      night  weekends   125.933333 14  07:00:00    morning  weekdays   447.200000 15  07:30:00    morning  weekends   545.200000 16  08:00:00    morning  weekends   668.733333 17  08:30:00    morning  weekends   751.333333 18  09:00:00    morning  weekdays   793.800000 19  09:30:00    morning  weekdays   781.125000 23  11:30:00       noon  weekdays   776.375000 24  12:00:00       noon  weekdays   741.812500 25  12:30:00       noon  weekends   723.000000 26  13:00:00       noon  weekends   734.562500 27  13:30:00       noon  weekends   763.882353 28  14:00:00  afternoon  weekdays   810.411765 31  15:30:00  afternoon  weekdays   855.411765 32  16:00:00  afternoon  weekdays   824.882353 33  16:30:00  afternoon  weekends   768.529412 34  17:00:00  afternoon  weekends   790.812500 35  17:30:00  afternoon  weekends   809.125000 

i want produce faceted grid of bar plots, 1 each variable (weekdays/weekends) , colour bars according "segment" column.

producing 2 bar plots straightforward:

g = sns.facetgrid(melted, col="variable") g.map(sns.barplot,'interval','value') 

this produces (i know xlabels wrong, can correct that):

i'm stuck @ colouring bars according "segment". per docs, need add variable when instantiating facetgrid , set palette:

g = sns.facetgrid(melted, col="variable",hue="segment",palette="set3") g.map(sns.barplot,'interval','value') 

but produces:

the bars stacked in front of each other instead of being spread across whole interval. missing here?

i've created gist dataset.

because interval nested within x variable (segment), need tell barplot of possible levels of x variable, not drawn on top of each other:

times = df.interval.unique() g = sns.facetgrid(df, row="variable", hue="segment",                   palette="set3", size=4, aspect=2) g.map(sns.barplot, 'interval', 'value', order=times)