https://prefect.io logo
Powered by
Title
m

matta

05/13/2020, 11:20 PM
What's the best way to pass an object containing credentials? I'm making an ETL to get stuff out of Google Sheets, and I'm using the 
gspread
package, which has you do everything from method calls to an authentication object. So like you go 
gc = gspread.service_account(filename=<filename>)
and point it at a special credentials file, then everything is through that. Should I just pass it to Secrets? Is there any risk of sensitive credentials being cached somewhere if I define the 
gc
object within the Flow itself?
n

nicholas

05/13/2020, 11:24 PM
Hi @matta! Secret tasks are exactly what you want to use for this; they're explicitly configured to never store their contents anywhere.
j

Jeremiah

05/13/2020, 11:29 PM
Hi @matta - I know a few folks who would love to see a Google Sheets task in the task library; if you come up with something you can share, please do!
:upvote: 1
m

matta

05/13/2020, 11:31 PM
Sweet, thanks, messing with that right now - if I find something reliable I'll hit you up! Might need some guidance for a proper Pull Request though.
@nicholas Thanks, took me a sec but I made a little version that works. 
class AuthenticateGsheets(SecretBase):
    def __init__(self, credentials_filename: Union[str, pathlib.Path], **kwargs: Any):
        self.credentials_filename = credentials_filename
        super().__init__(**kwargs)

    def run(self) -> gspread.client.Client:
        return gspread.service_account(filename=self.credentials_filename)
👍 1
🚀 1
j

Jeremiah

05/14/2020, 11:35 AM
Of course! We're happy to help you get it over the finish line
m

matta

05/16/2020, 1:10 AM
Just made a first draft. Not sure what to do next to get it into ship shape. https://github.com/mattalhonte/prefect/tree/gsheets/src/prefect/tasks/gsheets
Also not quite sure how to write tests for this, since it needs to hit Google Sheets.
j

Jeremiah

05/16/2020, 1:55 AM
Awesome! This is great
m

matta

05/17/2020, 11:48 PM
So is there a way to "wrap" all an object's methods with Prefect tasks? This has a bunch of little ones: https://gspread.readthedocs.io/en/latest/user-guide.html
Or maybe I should make a sort of "macro"? Something like 
@task
def gsheet_macro(worsksheet: gspread.models.Worksheet, fn: Callable):
    return Callable(worksheed)

gsheet_macro(gc, lambda x: x.find("Dough"))
@nicholas @Jeremiah
j

Jeremiah

05/18/2020, 11:25 PM
Hey @matta, sorry for the delay! That pattern might work, if the 
worksheet
object is Pickleable — but often, connection clients aren’t
Instead, you might have to create / instantiate the object in each task, then call the appropriate method. You could use a helper function to replace that boilerplate code inside the task, but I’d suggest not passing the gsheet client to the task directly.
m

matta

05/18/2020, 11:30 PM
Right on, thanks! So something like 
@task
def gsheet_macro(
    credentials_filename: Union[str, pathlib.Path] = None,
    sheet_key: str = None,
    worksheet_name: str = None,
    fn: Callable,
):
    client = AuthenticateGsheets(credentials_filename).run()
    google_sheet = client.open_by_key(sheet_key)
    worksheet = google_sheet.worksheet(worksheet_name)
    return Callable(worksheet)
Is it okay if I'm explicitly calling 
.run()
inside another task? Does that mess with Prefect in some way?
j

Jeremiah

05/18/2020, 11:42 PM
Generally speaking, it should be fine! But I think setup might not be quite what you want - I’d remove the 
@task
decorator at the top and instead make this just a function that returns tasks
actually never mind
I think for this pattern you’ll want a factory function which itself returns a 
@task
decorated function. In your current setup, where 
gsheet_macro
is decorated, 
fn
is going to be an input to your task (which seems unecessary). In addition, does 
AuthenticateGsheets
still need to be a task? maybe it can just be a helper function now?
def gsheet_helper(fn):
    @task
    def inner(credentials, sheet_key, name):
        client = ...
        return fn(...)
    return inner
^ something like that
But again, that’s just to cut down on the boilerplate of needing to instantiate and authenticate a client each time just to call one of its methods, there are many many ways you could approach this
m

matta

05/18/2020, 11:50 PM
AuthenticateGsheets
is a task right now to take advantage of subclassing 
SecretBase
as the "Prefectonic" way of handling authenticator objects. Not married to it if there's a better way!
j

Jeremiah

05/19/2020, 12:05 AM
I see — mainly that’s to help people ensure their credentials don’t accidentally get serialized as the results of a task, so I think there won’t be any issue with converting it to a function. HOWEVER, if you want to leave it a task, calling 
.run()
should be totally fine!
m

matta

05/19/2020, 12:07 AM
Ah, okay, I wasn't sure what got cached by Prefect
I think I'll leave it around for usage with the 
WriteGsheetRow
and `ReadGsheeRow`` tasks
Cool, thanks!
So once I've got that (with Docstrings and everything), what should I do to get it ready for the Pull Request? Not sure how to write Tests for it since it'd need to be hitting a Google Sheet.
j

Jeremiah

05/19/2020, 12:41 AM
Great! You can write tests for the non-google sheets parts. Have you used mocks to write tests for third party APIs before? If not it’s ok, you can open a PR with tests for what you can
In fact @Laura Lorenz (she/her) is working out a plan for the new 
contrib/
folder, and this would be a great candidate - once it has full tests (from you, Prefect, or anyone in the community) it can graduate to the core!
m

matta

05/19/2020, 12:47 AM
I haven't, actually. And cool!
Excited 🙂
maybe I'll give a try, though! Is there a library you'd recommend for mocking tests?
l

Laura Lorenz (she/her)

05/19/2020, 1:40 AM
Hi @matta, as Jeremiah said you can skip the tests and PR it into 
/src/prefect/contrib/tasks/{optionally your package name}/google_sheets.py
(or whatever you called your module) — and if you want to go for tests you can look at https://github.com/PrefectHQ/prefect/blob/master/tests/tasks/gcp/test_gcs_upload_download.py for some inspiration, the tests in the 
TestsBuckets
and 
TestBlob
classes show how to use unittest’s MagicMock and pytest.monkeypatch so that your tests don’t actually have to communicate with the google sheets API!
m

matta

05/19/2020, 1:54 AM
Awesome, thanks!
I'll try to clean it up a little bit and submit it tomorrow!
😄 2
Just submitted the PR!
😍 1
🚀 2
n

nicholas

05/20/2020, 2:26 AM
thanks @matta!
2 Views
#prefect-communityJoin Slack
Powered by