[ENH] Convert stock symbol to company name

[chrisfs]

Brief Description

I was thinking of a function that would either convert stock symbols to company names or create a new column of company names based on stock symbols from a different column, The only issue would be to find a stable site or api that would provide the company names in return for stock symbols and handling errors when you weren't connected to the internet.

Example API

def stock_to_name(df, symbol_column_name):

Before

Stock
F
O
GE

Stock2Name(DF.Stock)
Stock
Ford
Realty Income
General Motors.

[ericmjl]

There's definitely a bunch of APIs!

I've got very little domain knowledge on finance to know which one of these would be most reputable. Do you have an opinion?

[chrisfs]

I don't have much on the ground experience. If I had to make a decision right now, I would choose Bloomberg because the company is big and been around for a while. I don;t know anything about the API though.
I sent you an email through shortwhale about a related issue.

[hectormz]

Would it be possible to find a solution that generally didn't require an internet connection? Since it's not pulling stock quotes, and the dictionary of stock symbols is mostly stable, it'd be nice not to send a request each time.

[ericmjl]

In this case, I think caching the dictionary locally would be ideal. Doesn't have to be complicated: JSON/YAML or a sqlite database. (Actually, sqlite would be overkill, IMO!)

[szuckerman]

After a brief Google search it looks like there’s only about 4000 publicly traded companies on the NYSE/Nasdaq.

Wouldn’t it make sense to just store this in a Pandas DataFrame instead of json or YAML? Especially cause YAML has security vulnerabilities: https://tools.cisco.com/security/center/viewAlert.x?alertId=58783

[ericmjl]

That could work too! So we'd cache the pandas DataFrame, then? I think using the feather format for caching would be awesome here, especially since it's a binary format.

[hectormz]

Are you proposing to build this and distribute with pyjanitor, or it's built on the user's end, with the ability to rebuild at will?

[ericmjl]

I think the latter is the better way to approach this.

[szuckerman]

Just a thought on how to extend this:

Coming from an R background, R has a number of included datasets that are readable as a data.frame.

Similarly, scikit-learn has a number of included datasets, but those are included as arrays but not Pandas DataFrames.

Maybe for the submodules there could be added “informational” DataFrames? Obviously the stock names for finance, but maybe a periodic table for chemistry (I have no idea if that would be necessary) and similar ones for the others?

[ericmjl]

Ahhh, I see! Yes, we could also definitely make this part of the build pipeline, i.e. including scripts to build those dataframes as part of the CI or deployment story.

Gives me more impetus to do the auto-deployment thing right!

[hectormz]

I like that idea @szuckerman ! What would be some other general purpose DataFrames?