QST: Should the absence of tzdata package affect the performance in any way ?

[sdg002]

Research

• I have searched the [pandas] tag on StackOverflow for similar questions.

• I have asked my usage related question on StackOverflow.

Link to question on StackOverflow

https://stackoverflow.com/search?page=3&tab=Relevance&pagesize=30&q=pandas%20AND%20tzdata%20&searchOn=3

Question about pandas

We are on pandas 1.5.3. We are investigating some performance bottlenecks. At this point, we are not sure where the problem lies. However, we have a consistent pattern of observations.

We noticed that when tzdata==2025.2 was uninstalled, there was a severe degradation in performance (> 10x).

Upon further investigations and eliminations, we arrived at the following matrix:

Good perf-1

pandas==2.2.3
tzdata==2025.2

Good perf-2

pandas==1.5.3
tzdata==2025.2

Bad perf

No tzdata

pandas==1.5.3

Any suggestions ?

Is there any logic in any part of Pandas that relies on tzdata ?

Thanks,
Sau

[sdg002]

Sorry. SFO will not allow me to ask any more questions.

[mroeschke]

tzdata is used as an alternative timezone library compared to pytz. It's hard to answer performance implications without knowing what they are.

Note that pandas 1.5 is no longer supported and tzdata is a required dependency as of pandas 2.0

[sdg002]

Hello @mroeschke ,

Thanks for the quick reply. Based on your response, We have narrowed down our scenarios to the following:

Good - pandas 2.2.3 and tzdata

When pandas 2.2.3 is installed, tzdata gets installed too (tallies with what you have commented). This is our best case and everything works fine.

Bad - pandas 2.2.3 only

However, what surprised us that when tzdata is uninstalled using pip uninstall , the code continues to run without any errors. But, the performance is 20X slower.

We were expecting an error to be thrown.

It would be very helpful, if you can explain what pandas 2.2.3 does internally when the tzdata package is missing ?

Thanks,
Sau

[jbrockmendel]

explain what pandas 2.2.3 does internally when the tzdata package is missing

Well, it should raise at import-time since there's a check for it in pandas/__init__.py

@mroeschke i see it is also listed in compat/_optional. Maybe should be removed there?

[mroeschke]

Maybe should be removed there?

Yes definitely

[mroeschke]

Closing as I don't think there is anything actionable in this issue