You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Oct 30, 2022. It is now read-only.
Ensuring that we don't forget about user experience, perhaps by adding PR template with notes about the docs.
For meeshkan record:
Some questions that came up for me: If I already have a jsonl file and I go back to record more, does it update or overwrite? Is there any way to overwrite a specific file without deleting and regenerating? These questions could be solved via documentation or code, depending on how the product acts now.
One of the first logs says when running says spec generation mode disabled and I was like uhhhh what is this, is it bad 😆 Could be good to clarify for users what this means.
It currently doesn't tell you that it has created two directories (logs and specs)
Once the initial logs have printed, that's it. There's no helpful logs to reassure you that it's doing anything.
If you're using curl or another command-line tool to make requests, it should be done in a separate window because Meeshkan record is a blocking process. We don't mention this anywhere, so it's hard to know that this is what you're supposed to do. (Documented with Docs: Polish initial impression of running Meeshkan #144)
How many API calls are "enough"? For the tutorial, we do 33 requests to the Pokemon API - but in general we haven't stress tested this. Is it just the more calls, the more accurate? Or is there a minimum number of calls? Would be good to know.
How do you kill the recorder without losing data? Turns out you do ctrl + c or a kill command - but that's not written anywhere. It'd be nice to have that documented or, even better, have a stop command (Documented in Docs: Polish initial impression of running Meeshkan #144, but still in favor of a stop command)
Continuously build a schema when recording, and print a line every time the schema is updated due to a recording. That would give an indication when updates are slowing down/stopping, so the further recordings are not needed.
When creating my mock server, I needed to search for the servers field in OpenAPI spec to find the URL for the mock server to respond to. But I only knew to do that because Mike told me to. We should clarify in docs that this is located there and that it will usually be the same url as the API itself, unless you hack it. We haven’t mentioned hackability much in the existing documentation. If we write out the endpoints at startup, then this is less of a problem. See also the next note about logging endpoints at startup. Solved by Log endpoints at startup #142.
We are currently using http://localhost:9000/http://myapi.com/path/to/api, perhaps use http://localhost:9000/path/to/api without the host (that's what prism and mocklab does at least).
How do we know that it’s not calling the real API? The host URL is the same. Mock mode never calls the real API - should be specified.
Should we recommend that people run a virtual environment? I've always had a better time when I've used one and we recommend it in many of our tutorials.
Added a bunch of suggestions under four different categories: For meeshkan record, For meeshkan build, For meeshkan mock and General documentation improvements.
There are still two bugs that I need to file which will also potentially inform this list. One is related the differences between build modes (gen and replay) and the other is related to meeshkan record recognizing integers in query params.
This is a collection of tasks that can be done for improving the initial impressions of running meeshkan.
Some possible ones:
meeshkan --help
andmeeshkan CMD --help
and see if things can be improved (Improve --help output #124).For
meeshkan record
:spec generation mode disabled
and I was like uhhhh what is this, is it bad 😆 Could be good to clarify for users what this means.logs
andspecs
)ctrl + c
or akill
command - but that's not written anywhere. It'd be nice to have that documented or, even better, have astop
command (Documented in Docs: Polish initial impression of running Meeshkan #144, but still in favor of astop
command)For
meeshkan build
:-o
and-a
) (Docs: Polish initial impression of running Meeshkan #144)For
meeshkan mock
:-s
flag is doing in mock command example in MOCK.md (In meeshkan/meeshkan@608524e we write out--specs-dir
)servers
field in OpenAPI spec to find the URL for the mock server to respond to. But I only knew to do that because Mike told me to. We should clarify in docs that this is located there and that it will usually be the same url as the API itself, unless you hack it. We haven’t mentioned hackability much in the existing documentation. If we write out the endpoints at startup, then this is less of a problem. See also the next note about logging endpoints at startup. Solved by Log endpoints at startup #142.http://localhost:9000/http://myapi.com/path/to/api
, perhaps usehttp://localhost:9000/path/to/api
without the host (that's what prism and mocklab does at least).specs
. Or remove the default completely (Removed default in Simplify and enhance mock specs handling #154)--specs-dir
as an option, and instead take it as an argument. So run it likemeeshkan mock directory
instead ofmeeshkan mock --specs-dir directory
(Simplify and enhance mock specs handling #154) (documented in Docs: Polish initial impression of running Meeshkan #144)meeshkan mock file-or-directory
(Simplify and enhance mock specs handling #154)General documentation improvements:
--admin-port
instead of-a
) in documentation (Docs: Polish initial impression of running Meeshkan #144)--help
output.When working on this, let's create links to PR:s and concrete issues to visualise what has been done and ideas for further improvements.
The text was updated successfully, but these errors were encountered: