Read in CIK as string instead of int in the 10K metadata

[katie-lamb]

I'm in the process of creating a training dataset and am realizing that it would be nice to have a primary key for the SEC 10K filing archive that refers to each filing uniquely. It seems like CIK is just the ID for a filing company, not a primary key for the filings. A few questions:

• In the SEC 10K filing itself the CIK can have leading zeroes (e.g. 0000107815), but in the metadata the CIK is read in as an integer with GCSArchive().get_metadata(). CIK should probably be a string in the metadata? This might avoid conflicting CIKs for companies with very different names (see below)

• Right now there are multiple records in the metadata when the Company Name and Date Filed are the same but the filename is different. Does it make sense that there are multiple files for a company and filing date? Example: CIK = 1405332 has duplicate filings for 2016q1
• It could work to combine CIK and year quarter to create a primary key for each filing, but that won't work if a company files multiple 10K's per year quarter.
• Within the filing there's an SEC File Number which might be unique to the filing, but I'm not sure we can get that number without extracting the basic info at the top of the HTML filing.

[katie-lamb]

In the EDGAR database there may be duplicate filings for a company in the same year quarter and this is not a problem with the archiver (because maybe they've refiled or resubmitted). What we've done with FERC is take the most recent filing for each company and year quarter, this might be what we do for the 10K's too.

Seems like the only real issue is to make the CIK a string instead of an int.