kee
November 29, 2022, 10:48pm
#1
This issue on GitHub describes a bug within streamlit (likely) that dramatically limits the map rendering performance of pydeck objects passed into streamlit using local data. If you use pydeck this very much might be relevant to you.
opened 08:38PM - 12 Oct 22 UTC
type:bug
status:confirmed
priority:P2
feature:st.pydeck_chart
[](https://issues.streamlitapp.com/?issue=gh-5532)
TLDR - Pydeck when used with Streamlit has poor performance when passing in local data (vs. external URL). These issues are not present with pydeck standalone.
Iβve noticed dramatically different performance when using PyDeck standalone through a jupyter notebook vs. within a streamlit app. When the data object passed to a PyDeck layer is a local array or dataframe the performance is substantially poorer (itβs fine with an external URL, but thatβs not suitable for my use-case).
Note: I am using streamlit v1.13 and pydeck 0.8.3b (with geojson layers).
For transparency I posted this here without any responses: https://discuss.streamlit.io/t/pydeck-performance-issues-with-streamlit-but-not-standalone/31434
But think that this is a legitimate issue as it was brought up in [Jan 2021](https://github.com/streamlit/streamlit/issues/2569). While that issue is closed, there was no resolution - and the suggestion to render pydeck elements as raw embedded HTML e.g. via `components.html(deck.to_html(as_string=True), height=600)` does improve performance but has other issues associated with it.
---
Community voting on feature requests enables the Streamlit team to understand which issues are most important to our users.
**If you'd like the Streamlit team to prioritize this issue, please use the π (thumbs up emoji) reaction in response to the initial post.**

Please +1 the on the first comment or add to the discussion.
Thanks!