It definitely works on DocumentClient.upload. I haven't used upload_directory, but that method uses upload, so it should work.

There is some inconsistency with client.documents.upload and the project argument. When uploading a PDF file from your computer you can only use a string in the project argument, but when you link to a PDF URL it accepts both a string and an integer. It took me some time to understand why uploading did not work anymore when I switched from using URLs to local files, so fixing this might avoid some frustration for others :D

Sorry for this confusion. I'll take a look at it. Though could you please clarify what you mean when you say "link to a PDF url."

Thanks!
To clarify, see https://python-documentcloud.readthedocs.org/en/latest/documents.html#uploading
It says: You can also URLs link to PDFs, if that’s the kind of thing you want to do.
client.documents.upload("http://ord.legistar.com/Chicago/attachments/e3a0cbcb-044d-4ec3-9848-23c5692b1943.pdf")

Weird. Now that's something I hadn't considered.

Regardless of how you submit a document, the project keyword argument is passed along in the same format by this bit of code.

The one differnece I see that is URL uploads do not use a "multipart" post handler to pass along all that PDF data. Though I'm not sure that should make much difference.

Can you provide some examples of the code you're running that's not working? Are there error messages?

I've been running into an ish when I try to upload to a specific project. Here's the function I'm calling:

from documentcloud import DocumentCloud
import os


def dc_upload_handler(fileobj, title, description, proj_name, access_level):
    dc_username = os.environ['DC_USERNAME']
    dc_password = os.environ['DC_PASSWORD']
    client = DocumentCloud(dc_username, dc_password)
    obj = client.documents.upload(
        fileobj,
        title=title,
        description=description,
        project=proj_name,
        access=access_level
    )
    return client.documents.get(obj.id)

I'm not getting any errors, and the document uploads fine -- it's just not attached to the project I specified. I've confirmed that the string I'm passing for the project matches the name of the project on DC. I also tried fetching the project first -- client.projects.get_by_title(proj_name) -- and then passing that as the parameter to the upload function, but no dice.

What are you passing in as the proj_name variable?

The project name as a string. E.g.: dc_upload_handler("test.pf", "Title", "Description of PDF", "Project Name", "public")

According to the official API documentation, DocumentCloud expected the "numeric Project id" for that input. Try that and see if that works.

Bingo. Thanks.

This is something we should spell out more clearly in the documentation. I'm curious where you looked that failed to make this need clear.

Documents >> uploading. My though process was: if you can fetch a project by either its ID or title, maybe you could also post to one in the same way. (I might add a few lines to my function to fetch the project by name and return the ID.)

yeah that's a pretty reasonable thing to think. I wonder if we should be testing if an object is submitted, and if so pulling the ID.