Investigate serializing all plots before/at startup to speed up app

[alyssadai]

Per question (subquestion="all"), we need 6 variations of graphs:

1. Per state, we need 2 plots:
   1. one with threshold=None (showing all endorsement/response levels)
   2. one with threshold=DEFAULT_QUESTION["outcome"] (thresholding responses at "3+")
   3. For national-level responses (state=None), we need 4 plots:
   4. one with threshold=None (showing all endorsement/response levels)
   5. one with threshold=DEFAULT_QUESTION["outcome"] (thresholding responses at "3+")
   6. one with threshold=None AND stratify=True (stratify bar graph by party affiliation)
   7. one with threshold=DEFAULT_QUESTION["outcome"] AND stratify=True

List of states can be obtained from one of the following:

• state column of DATA_DICTIONARIES["state_abbreviations.tsv"]
• state column of SURVEY_DATA["samplesizes_state.tsv"]

List of questions can be obtained from:

• question column of DATA_DICTIONARIES["question_dictionary.tsv"]

Where should the figures live?

Maybe under code/assets/, and the script can live in code/?

[alyssadai]

Hey @surchs, thanks for helping to tackle this issue!

I've updated the iss description with more details of the total variants of graphs needed, and where you can find the necessary question/state info. Let me know if I missed anything!

[surchs]

So the good news is: serialization works nicely. The bad news is: loading the serialized object into memory takes around 45 seconds. We'll have to see how often the dashboard will actually do that.