Can't get correct dim in link_selections instance

I tired to create an interactive plot between two dimensions by using two separated histograms. However, once the link_selections.instance() was created and added necessary plots, the box_select didn’t work correctly and there was nothing happened on two charts neither.

The code looks like below

import hvplot.pandas
# and other necessary libraries

pd.options.plotting.backend = 'holoviews'

df = pd.read_pickle(file_path)

# I plan to make hist chart on 'speed_avg' and 'enginespeed_avg' columns with int32 dtype

# the whole dataset contains over 5 million items
# I just use the first 500 rows as an example
a = df['speed_avg'].loc[:500].hvplot.hist(bins=20).options(width=600, tools=['tap', 'box_select', 'hover'])
b = df['enginespeed_avg'].loc[:500].hvplot.hist(bins=30).options(width=600, xlim=(0, 3000), tools=['tap', 'box_select', 'hover'])

ls = hv.link_selections.instance()
ls(a) + ls(b)

such code above will show two hist charts correctly in my jupyter notebook, but nothing happened when I selected some part of a neither b.

I then ran ls.selection_expr and it returned (dim('enginespeed_avg')>=830.58507391747)&(dim('enginespeed_avg')<=1378.457414343002) which looked good.

But when I eliminated the jupyer notebook and went back the to the script, it will show an error info below such like: linked_selection aborted because it could not display selection for all elements: One or more dimensions in the expression (dim('enginespeed_avg')>=83058507391747)&(dim('enginespeed_avg')<=1378457414343002) could not resolve on ':Dataset [index] (speed_avg)' Ensure all dimensions referenced by the expression are present on the supplied object on ':Histogram [speed_avg] (speed_avg_count)'.

Notice that it returns a wrong dim range 83058507391747 and 1378457414343002. Comparing with what we got in expr order 830.58507391747 and 1378.457414343002, it just transform a float data to a huge int without the point

Not quite sure if others got the same issue, or maybe some other best practice to avoid this case?
Many thanks for your support!

Welcome to the community!

Can you try to make a minimal, reproducible example (MRE)?

The pickle file you used is not available, and even if it was I would not load it on my computer because it is unsafe. Try to use another file format instead like parquet or csv. And try to reduce it from 5 million items to 100ish (if possible).