This section includes guides for advanced usage patterns.
To add your own custom location handler, write a function that receives a request, and a Schema <marshmallow.Schema>
, then decorate that function with Parser.location_loader <webargs.core.Parser.location_loader>
.
from webargs import fields
from webargs.flaskparser import parser
@parser.location_loader("data")
def load_data(request, schema):
return request.data
# Now 'data' can be specified as a location
@parser.use_args({"per_page": fields.Int()}, location="data")
def posts(args):
return "displaying {} posts".format(args["per_page"])
Note
The schema is passed so that it can be used to wrap multidict types and unpack List fields correctly. If you are writing a loader for a multidict type, consider looking at MultiDictProxy <webargs.multidictproxy.MultiDictProxy>
for an example of how to do this.
You can define your own locations which mix data from several existing locations.
The json_or_form location does this -- first trying to load data as JSON and then falling back to a form body -- and its implementation is quite simple:
def load_json_or_form(self, req, schema):
"""Load data from a request, accepting either JSON or form-encoded
data.
The data will first be loaded as JSON, and, if that fails, it will be
loaded as a form post.
"""
data = self.load_json(req, schema)
if data is not missing:
return data
return self.load_form(req, schema)
You can imagine your own locations with custom behaviors like this. For example, to mix query parameters and form body data, you might write the following:
from webargs import fields
from webargs.multidictproxy import MultiDictProxy
from webargs.flaskparser import parser
@parser.location_loader("query_and_form")
def load_data(request, schema):
# relies on the Flask (werkzeug) MultiDict type's implementation of
# these methods, but when you're extending webargs, you may know things
# about your framework of choice
newdata = request.args.copy()
newdata.update(request.form)
return MultiDictProxy(newdata, schema)
# Now 'query_and_form' means you can send these values in either location,
# and they will be *mixed* together into a new dict to pass to your schema
@parser.use_args({"favorite_food": fields.String()}, location="query_and_form")
def set_favorite_food(args):
... # do stuff
return "your favorite food is now set to {}".format(args["favorite_food"])
When you need more flexibility in defining input schemas, you can pass a marshmallow Schema <marshmallow.Schema> instead of a dictionary to Parser.parse <webargs.core.Parser.parse>, Parser.use_args <webargs.core.Parser.use_args>, and Parser.use_kwargs <webargs.core.Parser.use_kwargs>.
from marshmallow import Schema, fields
from webargs.flaskparser import use_args
class UserSchema(Schema):
id = fields.Int(dump_only=True) # read-only (won't be parsed by webargs)
username = fields.Str(required=True)
password = fields.Str(load_only=True) # write-only
first_name = fields.Str(missing="")
last_name = fields.Str(missing="")
date_registered = fields.DateTime(dump_only=True)
# NOTE: Uncomment below two lines if you're using marshmallow 2
# class Meta:
# strict = True
@use_args(UserSchema())
def profile_view(args):
username = args["userame"]
# ...
@use_kwargs(UserSchema())
def profile_update(username, password, first_name, last_name):
update_profile(username, password, first_name, last_name)
# ...
# You can add additional parameters
@use_kwargs({"posts_per_page": fields.Int(missing=10)}, location="query")
@use_args(UserSchema())
def profile_posts(args, posts_per_page):
username = args["username"]
# ...
Warning
If you're using marshmallow 2, you should always set strict=True
(either as a class Meta
option or in the Schema's constructor) when passing a schema to webargs. This will ensure that the parser's error handler is invoked when expected.
Any Schema <marshmallow.Schema> passed to use_kwargs <webargs.core.Parser.use_kwargs> MUST deserialize to a dictionary of data. If your schema has a post_load <marshmallow.decorators.post_load> method that returns a non-dictionary, you should use use_args <webargs.core.Parser.use_args> instead.
from marshmallow import Schema, fields, post_load
from webargs.flaskparser import use_args
class Rectangle:
def __init__(self, length, width):
self.length = length
self.width = width
class RectangleSchema(Schema):
length = fields.Float()
width = fields.Float()
@post_load
def make_object(self, data, **kwargs):
return Rectangle(**data)
@use_args(RectangleSchema)
def post(self, rect: Rectangle):
return f"Area: {rect.length * rect.width}"
Packages such as marshmallow-sqlalchemy and marshmallow-dataclass generate schemas that deserialize to non-dictionary objects. Therefore, use_args <webargs.core.Parser.use_args> should be used with those schemas.
If you need to parametrize a schema based on a given request, you can use a "Schema factory": a callable that receives the current request and returns a marshmallow.Schema instance.
Consider the following use cases:
- Filtering via a query parameter by passing
only
to the Schema. - Handle partial updates for PATCH requests using marshmallow's partial loading API.
from flask import Flask
from marshmallow import Schema, fields
from webargs.flaskparser import use_args
app = Flask(__name__)
class UserSchema(Schema):
id = fields.Int(dump_only=True)
username = fields.Str(required=True)
password = fields.Str(load_only=True)
first_name = fields.Str(missing="")
last_name = fields.Str(missing="")
date_registered = fields.DateTime(dump_only=True)
def make_user_schema(request):
# Filter based on 'fields' query parameter
fields = request.args.get("fields", None)
only = fields.split(",") if fields else None
# Respect partial updates for PATCH requests
partial = request.method == "PATCH"
# Add current request to the schema's context
return UserSchema(only=only, partial=partial, context={"request": request})
# Pass the factory to .parse, .use_args, or .use_kwargs
@app.route("/profile/", methods=["GET", "POST", "PATCH"])
@use_args(make_user_schema)
def profile_view(args):
username = args.get("username")
# ...
We can reduce boilerplate and improve [re]usability with a simple helper function:
from webargs.flaskparser import use_args
def use_args_with(schema_cls, schema_kwargs=None, **kwargs):
schema_kwargs = schema_kwargs or {}
def factory(request):
# Filter based on 'fields' query parameter
only = request.args.get("fields", None)
# Respect partial updates for PATCH requests
partial = request.method == "PATCH"
return schema_cls(
only=only, partial=partial, context={"request": request}, **schema_kwargs
)
return use_args(factory, **kwargs)
Now we can attach input schemas to our view functions like so:
@use_args_with(UserSchema)
def profile_view(args):
# ...
get_profile(**args)
See the "Custom Fields" section of the marshmallow docs for a detailed guide on defining custom fields which you can pass to webargs parsers: https://marshmallow.readthedocs.io/en/latest/custom_fields.html.
Using the Method <marshmallow.fields.Method>
and Function <marshmallow.fields.Function>
fields requires that you pass the deserialize
parameter.
@use_args({"cube": fields.Function(deserialize=lambda x: int(x) ** 3)})
def math_view(args):
cube = args["cube"]
# ...
To add your own parser, extend Parser <webargs.core.Parser>
and implement the load_* method(s) you need to override. For example, here is a custom Flask parser that handles nested query string arguments.
import re
from webargs import core
from webargs.flaskparser import FlaskParser
class NestedQueryFlaskParser(FlaskParser):
"""Parses nested query args
This parser handles nested query args. It expects nested levels
delimited by a period and then deserializes the query args into a
nested dict.
For example, the URL query params `?name.first=John&name.last=Boone`
will yield the following dict:
{
'name': {
'first': 'John',
'last': 'Boone',
}
}
"""
def load_querystring(self, req, schema):
return _structure_dict(req.args)
def _structure_dict(dict_):
def structure_dict_pair(r, key, value):
m = re.match(r"(\w+)\.(.*)", key)
if m:
if r.get(m.group(1)) is None:
r[m.group(1)] = {}
structure_dict_pair(r[m.group(1)], m.group(2), value)
else:
r[key] = value
r = {}
for k, v in dict_.items():
structure_dict_pair(r, k, v)
return r
If you'd prefer validation errors to return status code 400
instead of 422
, you can override DEFAULT_VALIDATION_STATUS
on a Parser <webargs.core.Parser>
.
from webargs.falconparser import FalconParser
class Parser(FalconParser):
DEFAULT_VALIDATION_STATUS = 400
parser = Parser()
use_args = parser.use_args
use_kwargs = parser.use_kwargs
In order to parse a JSON array of objects, pass many=True
to your input Schema
.
For example, you might implement JSON PATCH according to RFC 6902 like so:
from webargs import fields
from webargs.flaskparser import use_args
from marshmallow import Schema, validate
class PatchSchema(Schema):
op = fields.Str(
required=True,
validate=validate.OneOf(["add", "remove", "replace", "move", "copy"]),
)
path = fields.Str(required=True)
value = fields.Str(required=True)
@app.route("/profile/", methods=["patch"])
@use_args(PatchSchema(many=True))
def patch_blog(args):
"""Implements JSON Patch for the user profile
Example JSON body:
[
{"op": "replace", "path": "/email", "value": "mynewemail@test.org"}
]
"""
# ...
Arguments for different locations can be specified by passing location
to each use_args <webargs.core.Parser.use_args> call:
# "json" is the default, used explicitly below
@app.route("/stacked", methods=["POST"])
@use_args({"page": fields.Int(), "q": fields.Str()}, location="query")
@use_args({"name": fields.Str()}, location="json")
def viewfunc(query_parsed, json_parsed):
page = query_parsed["page"]
name = json_parsed["name"]
# ...
To reduce boilerplate, you could create shortcuts, like so:
import functools
query = functools.partial(use_args, location="query")
body = functools.partial(use_args, location="json")
@query({"page": fields.Int(), "q": fields.Int()})
@body({"name": fields.Str()})
def viewfunc(query_parsed, json_parsed):
page = query_parsed["page"]
name = json_parsed["name"]
# ...
- See the
Framework Support <framework_support>
page for framework-specific guides. - For example applications, check out the examples directory.