Codestin Search App

MarcoGorelli · 2026-02-25T11:36:00Z

Summary

As discussed, following on from the hackmd document (thanks @javabster for helpful comments!)

Fixes #XXXX

Test Plan

jorenham · 2026-02-25T13:25:30Z

website/blog/2026-02-25-typing-pandas.md

+In order to improve the developer experience for pandas' users across the ecosystem, we decided to focus on improving pandas' typing. Why? Because better type hints mean:
+
+- More accurate and useful auto-completions from VSCode / PyCharm / NeoVIM / Positron / other IDEs.
+- More robust pipelines, as some categories of bugs can be caught without even needing to execute your code.


Maybe also mention the (alleged) LLM benefits?

sure, thanks - did you have a reference in mind for this?

The closest I was able to find is https://www.se.cs.uni-saarland.de/conferences/ASE/ase2023/details/ase-2023/ase-2023-papers/12/Generative-Type-Inference-for-Python.html, but I don't think there's anything yet that tests this for modern LLMs.

There's also https://llm-guidelines.org/study-types/, which suggests that structured outputs (which if I understand correctly also includes static typing) is indeed helpful.

thanks - as far as I can tell, that paper's about using llms to do type inference? if so, not sure if we should cite it for the alleged llm benefits of having typed code

Do we need even need a citation? I mean; I'm all for being accurate, but in this case I doubt that anyone would question that static typing helps LLMs write better code, seeing as it also helps humans write better code 🤷‍♂️

But I'm assuming here that the types are correct. Because if not, I wouldn't be surprised that LLMs perform worse than if there are no types at all. The same holds for humans, after all.

tbh it's not obvious to me that they would perform better, they hallucincate method names all the time and i find that they often suggest code that which doesn't satisfy type-checkers even in codebases that are fully typed

i'd prefer to leave this out unless we have a reference if it's ok

website/blog/2026-02-25-typing-pandas.md

javabster · 2026-02-25T14:39:02Z

website/blog/2026-02-25-typing-pandas.md

+
+pandas is one of the most widely used Python libraries. At time of writing, it is [downloaded about half-a-billion times per month from PyPI](https://pypistats.org/packages/pandas), is supported by nearly all Python data science packages, and is generally required learning in data science curriculums. Despite modern alternatives existing, pandas' impact cannot be minimised or understated.
+
+In order to improve the developer experience for pandas' users across the ecosystem, we decided to focus on improving pandas' typing. Why? Because better type hints mean:


I think we should still be more explicit here about who "we" is at the beginning, could you add a clarification, even if its just briefly in brackets? Something like "the team at Quantsight" or "the Quantsight team with support from the Pyrefly team", whatever you feel is appropriate. My main concern is that people coming to the blog on the pyrefly website will assume "we" means just the Pyrefly team

javabster · 2026-02-25T14:42:02Z

website/blog/2026-02-25-typing-pandas.md

+
+## Beyond Pyright - what about "Pyrefly report"?
+
+Pyright's verifytypes feature takes about 2 and a half minutes to run in pandas-stubs. There's room of improvement here - so much so, that the Pyrefly team is working on a [`pyrefly report`](https://pyrefly.org/en/docs/report/) which would work similarly. The `pyrefly report` API is not yet considered stable, so for now pandas-stubs uses Pyright's `--verifytypes` command, but hopefully a faster is on the horizon!


formatting: should it be verifytypes? or verify types?

typo: "hopefully a faster is on the horizon!" a faster tool?

formatting: should it be verifytypes? or verify types?

--verifytypes is correct; pyright --help shows:

Usage: pyright [options] files... Options: [..] --verifytypes <PACKAGE> Verify type completeness of a py.typed package [..]

I meant the instance in the first sentence (Pyrights veriftypes feature...) not the --verifytypes one :)

javabster · 2026-02-25T14:46:23Z

website/blog/2026-02-25-typing-pandas.md

@@ -0,0 +1,76 @@
+---
+title: pandas' public API is now type-complete!


Suggested change

title: pandas' public API is now type-complete!

title: Pandas' Public API Is Now Type-Complete!

Please use title case for titles :)

they ask that it be used lowercase even at the beginning of a sentece https://pandas.pydata.org/about/citing.html#brand-and-logo

When using the project name pandas, please use it in lower case, even at the beginning of a sentence.

if we're ok going against that in titles, then sure, will do

oh! Thanks for flagging, lets follow their citation guidelines, but the rest of the title should still be title case imho

MarcoGorelli · 2026-02-25T15:35:35Z

thanks for your reviews! 🙏

javabster

LGTM 🚀

but lets wait to merge this until early next week, we already published a blog earlier this week

samwgoldman · 2026-02-27T20:24:36Z

website/blog/2026-02-25-typing-pandas.md

+- `DataFrame` is reported as "partially unknown" because its method `.index` returns `Index`, which is partially unknown.
+- `Index` is reported as "partially unknown" because its method `to_series` returns `Series`, which is partially unknown.
+- `Series` is reported as "partially unknown" because its method `to_frame` returns `DataFrame`, which is partially unknown.


This is surprising to me / doesn't make a ton of sense to me. DataFrame is unknown because DataFrame is unknown? I'd expect there to be some "Unknown" or "Any" typed attribute or similar.

i've reworked the example so it's clearer, thanks for commenting!

add blog post

c1e12f5

meta-cla bot added the cla signed label Feb 25, 2026

jorenham reviewed Feb 25, 2026

View reviewed changes

javabster reviewed Feb 25, 2026

View reviewed changes

titlecase, truncate, clarify "we"

2688eda

missing word

d9429f7

maggiemoss assigned javabster Feb 26, 2026

javabster approved these changes Feb 26, 2026

View reviewed changes

samwgoldman reviewed Feb 27, 2026

View reviewed changes

MarcoGorelli added 3 commits March 2, 2026 11:09

clarify the virality part

72dbda4

verifytypes formatting

60eff79

replace "here" hyperlink

43acb96


		pandas is one of the most widely used Python libraries. At time of writing, it is [downloaded about half-a-billion times per month from PyPI](https://pypistats.org/packages/pandas), is supported by nearly all Python data science packages, and is generally required learning in data science curriculums. Despite modern alternatives existing, pandas' impact cannot be minimised or understated.

		In order to improve the developer experience for pandas' users across the ecosystem, we decided to focus on improving pandas' typing. Why? Because better type hints mean:


		## Beyond Pyright - what about "Pyrefly report"?

		Pyright's verifytypes feature takes about 2 and a half minutes to run in pandas-stubs. There's room of improvement here - so much so, that the Pyrefly team is working on a [`pyrefly report`](https://pyrefly.org/en/docs/report/) which would work similarly. The `pyrefly report` API is not yet considered stable, so for now pandas-stubs uses Pyright's `--verifytypes` command, but hopefully a faster is on the horizon!

		@@ -0,0 +1,76 @@
		---
		title: pandas' public API is now type-complete!

Conversation

MarcoGorelli commented Feb 25, 2026

Summary

Test Plan

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jorenham Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jorenham Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

javabster Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

MarcoGorelli commented Feb 25, 2026

Uh oh!

javabster left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

jorenham Feb 25, 2026 •

edited

Loading

jorenham Feb 25, 2026 •

edited

Loading

javabster Feb 25, 2026 •

edited

Loading