Certain counttypes in queries--especially TextRatio and WordsRatio--can produce division by zero. (Say you're comparing publication rates by state before and after 1922, and some states do not exist in the post-1922 set.
This doesn't affect the main browser, but does affect my Vega-Lite experiments.
I believe the problem is probably that we are serializing to JSON with allow_nan to be false. That's good behavior, but python doesn't do what Javascript likes, which is to to replace NaN/Infinity values with 'null'. That would be one option. Another would be to just silently drop any queries that do not exist in the compare_limits field, perhaps as part of the join.
Certain counttypes in queries--especially
TextRatioandWordsRatio--can produce division by zero. (Say you're comparing publication rates by state before and after 1922, and some states do not exist in the post-1922 set.This doesn't affect the main browser, but does affect my Vega-Lite experiments.
I believe the problem is probably that we are serializing to JSON with
allow_nanto be false. That's good behavior, but python doesn't do what Javascript likes, which is to to replace NaN/Infinity values with 'null'. That would be one option. Another would be to just silently drop any queries that do not exist in the compare_limits field, perhaps as part of the join.