alanmitchell / fnsb-benchmark Goto Github PK

Check for internal consistency between different report sections, at a minimum.

The truncating is the largest problem. We could live with vertically stacking the graphs, although it would be better to have them shown as they are in the browser. Do we need to set an explicit Pixel Size to get them to fit horizontally and not wrap?

Colors are not the same between Usage and Cost Pie Charts.

UserWarning: Attempting to set identical bottom==top results

df.electricity + df.district_heat will produce NaNs for any rows where a NaN appears in the Series. One way to avoid this is to do:

df[['electricity', 'district_heat']].sum(axis=1)

NaNs will be treated as 0s with this method of addition, which is what we often want.

Another way, but more cumbersome:

df.electricity.fillna(0.0) + df.district_heat.fillna(0.0)

Fonts are too small to read. Also, maybe we can get the chart to fill more of the figure? Part of the problem is margins around the chart, but my discussion with Conroy made it sound like that may be difficult to fix.

DataFrame needs to be sorted ascending before applying the Pandas pct_change() method.

It may just be the Building Heating Unit Cost that is calculated wrong (too high).

Found a discrepancy for kW avg and kW max for the Big Dipper and Hamme Swimming Pool (DIPMP1 and HSPSWP)

Lathrop High School, for example. Could draw them at an angle, as one solution. Although not sure why on some buildings this is not a problem.

So that individual items do not wrap to multiple lines.

Show how to unpickle the raw and pre-processed DataFrames and how to call the functions in the benchmark.py module, if needed. Import bench_util and graph_util.

#if site_id < 'ASLSQD': continue

Seems like it would be more robust to say if the Units are kW, then it is a demand cost; for all other charges, it is an energy cost.

Maybe this is a Future feature, but I assume this gets rid of the Matplotlib warning about same top and bottom?

ASLC47 has Oil #1 data only

Will Need to skip portions of the code in analyze_site() function. Also, we need to make Jinja2 not raise an error when variables are missing. This page shows how to do that:
https://stackoverflow.com/questions/6190348/jinja2-ignore-undefinederrors-for-objects-that-arent-found