Open paddymul opened 8 months ago
For notes, I'm adding my thoughts on this that I sent to @MarcoGorelli
Looking through the dataframe standard, the next time I improve on the JS deserialization of dataframes, I'm going to implement the dataframe_standard in JS.
This will allow me to decouple the application side from the serialization side. Then I can change the core serialization. I'm specifically thinking of higher performance serializations based on TypedArrays and base64. I notice that the standard doesn't have a "get_row_by_id" or "get_as_list_of_dicts". I understand why those aren't important for numeric python/C programming, but they are very common in JS.
FWIW I have written multiple versions of Dataframe to JSON to JS serialization over the past decade. All jank, and slow. I want to build one properly, also the python side.
No promises on when I will build this, but I will use your code as the basis for it next time I do.
I would like to have better frontend dataframe manipulation for simple tasks.
[cleanedVal, origVal, annotation]
from raw columns, the combination done in JS. https://github.com/paddymul/buckaroo/issues/74I want to research the existing js dataframe like libraries. I don't think they offer this functionality, but I want to check.
Once the API is decided on and implemented, this library will enable performance increases through better serialization.