graphql-core-legacy
graphql-core-legacy copied to clipboard
Order of Middlewares
I've discovered that this line inverts the middleware order: https://github.com/graphql-python/graphql-core/blob/6df8a6312b579a6a1454bcf29a566ce5d0fa9849/graphql/execution/middleware.py#L51
This was somewhat confusing, as I expected the first installed middleware to be the entrypoint. I.e.
schema.execute(
statement,
middleware=[mw1, mw2],
)
actually produces the call stack mw2 -> mw1 -> resolve_func
, whereas i'd expected mw1 -> mw2 -> resolve_func
.
I'm not clear if this is a bug, as this functionality isn't tested anywhere, or if it would just go in documentation somewhere.
I confirm that the middleware's resolve()
methods get called in reverse order from how they're listed when passed into schema.execute
. Let's say your middleware is:
class CharPrintingMiddleware(object):
def __init__(self, char):
self.char = char
def resolve(self, next, *args, **kwargs):
print(f'resolve() called for middleware {self.char}')
return next(*args, **kwargs).then(
lambda x: print(f'then() for {self.char}')
)
and you call:
schema.execute(
statement,
middleware=[
CharAppendingMiddleware('a'),
CharAppendingMiddleware('b'),
CharAppendingMiddleware('c')
]
)
Then the printed output is:
resolve() called for middleware c
resolve() called for middleware b
resolve() called for middleware a
then() for a
then() for b
then() for c
This makes me wonder if perhaps it was intentional to have middleware called in reverse order, and the intention was that then()
set up by the middleware would wind up being called in order. @syrusakbary you git blame to those lines in middleware.py from back in 2016, do you recall if this is intentional or a minor, fixable bug?