This repository has been archived by the owner on Jul 3, 2023. It is now read-only.
Async implementation of driver/adapter #171
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
elijahbenizzy
force-pushed
the
async-prototype
branch
from
7908643 to
081db80
Compare
skrawcz
reviewed
Aug 6, 2022
skrawcz
reviewed
Aug 6, 2022
skrawcz
reviewed
Aug 6, 2022
elijahbenizzy
force-pushed
the
async-prototype
branch
from
081db80 to
de39759
Compare
skrawcz
reviewed
Aug 7, 2022
elijahbenizzy
force-pushed
the
async-prototype
branch
4 times, most recently
from
018fd96 to
a8667ca
Compare
elijahbenizzy
changed the title
elijahbenizzy
force-pushed
the
async-prototype
branch
from
a8667ca to
f5c9826
Compare
|
Remaining:
|
Basically we pass the coroutine the entire way through. When a function depends on other data, we create a coroutine that awaits everything. Finally, we do a gather at the end. This involves no modification to the functiongraph code
elijahbenizzy
force-pushed
the
async-prototype
branch
from
f5c9826 to
6910d33
Compare
skrawcz
reviewed
Aug 7, 2022
skrawcz
reviewed
Aug 7, 2022
elijahbenizzy
force-pushed
the
async-prototype
branch
from
6910d33 to
97967a8
Compare
If we use tasks, we can await them twice. Furthermore, they'll begin scheduling earlier, and python handles keeping track of them. This is a much cleaner way to handle it.
skrawcz
suggested changes
Aug 8, 2022
examples/async/fastapi_example.py
Outdated
Comment on lines
52
to
58
| ad_hoc_utils.create_temporary_module( | ||
| pipeline, | ||
| computation1, | ||
| foo, | ||
| bar, | ||
| some_data, | ||
| computation2)) |
There was a problem hiding this comment.
We should probably not do this in an example. This doesn't feel very "ad hoc".
Otherwise 💡, fastapi could be a possible code compilation target too.
hamilton/experimental/h_async.py
Outdated
| if display_graph: | ||
| raise ValueError(f'display_graph=True is not supported for the async graph adapter. ' | ||
| f'Instead you should be using visualize_execution.') | ||
| return await await_dict_of_tasks({key: asyncio.create_task(process_value(memoized_computation[key])) for key in final_vars}) |
There was a problem hiding this comment.
Suggested change
| return await await_dict_of_tasks({key: asyncio.create_task(process_value(memoized_computation[key])) for key in final_vars}) | |
| task_dict = {key: asyncio.create_task(process_value(memoized_computation[key])) for key in final_vars} | |
| return await await_dict_of_tasks(task_dict) |
Also is process_value needed here? Seems a bit redundant with the result being wrapped in a task and that awaited?
There was a problem hiding this comment.
process_value is needed but create_task is not I think...
There was a problem hiding this comment.
okay. But this seems like an important line that should be at least two... break out the dict comprehension at least.
skrawcz
reviewed
Aug 10, 2022
hamilton/experimental/h_async.py
Outdated
| callabl = node.callable | ||
|
|
||
| async def new_fn(fn=callabl, **fn_kwargs): | ||
| fn_kwargs = await await_dict_of_tasks({key: process_value(value) for key, value in fn_kwargs.items()}) |
There was a problem hiding this comment.
can move this dict comprehension to its own line.
skrawcz
approved these changes
Aug 10, 2022
elijahbenizzy
force-pushed
the
async-prototype
branch
2 times, most recently
from
ca4b36f to
bc3a7e2
Compare
1. Fixes docstring formats 2. Removes redundant call to create_task 3. Breaks example into modules
elijahbenizzy
force-pushed
the
async-prototype
branch
from
bc3a7e2 to
e044a40
Compare
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
@skrawcz this is what I meant
Basically we pass the coroutine the entire way through.
When a function depends on other data, we create a coroutine that
awaits everything. Finally, we do a gather at the end.
I think this is optimal but will need to dig in a bit/look into
things.
Otherwise, nifty POC with very rough edges.
The biggest hack is that I'm using a cache to store coroutine values as python makes it very hard to access them.
I think we can do better, but the point is we don't have to mess with execution.
[Short description explaining the high-level reason for the pull request]
Changes
Testing
Notes
Checklist
Testing checklist
Python - local testing