Beginning in v21.3, Sanic makes use of this new AST-style router in two use cases:
- Routing paths; and
- Routing signals.
Therefore, this package comes with a BaseRouter that needs to be subclassed in order to be used for its specific needs.
Most Sanic users should never need to concern themselves with the details here.
A simple implementation:
import logging
from sanic_routing import BaseRouter
logging.basicConfig(level=logging.DEBUG)
class Router(BaseRouter):
def get(self, path, *args, **kwargs):
return self.resolve(path, *args, **kwargs)
router = Router()
router.add("/<foo>", lambda: ...)
router.finalize()
router.tree.display()
logging.info(router.find_route_src)
route, handler, params = router.get("/matchme", method="BASE", extra=None)The above snippet uses router.tree.display() to show how the router has decided to arrange the routes into a tree. In this simple example:
<Node: level=0>
<Node: part=__dynamic__:str, level=1, groups=[<RouteGroup: path=<foo:str> len=1>], dynamic=True>
We can can see the code that the router has generated for us. It is available as a string at router.find_route_src.
def find_route(path, method, router, basket, extra):
parts = tuple(path[1:].split(router.delimiter))
num = len(parts)
# node=1 // part=__dynamic__:str
if num == 1: # CHECK 1
try:
basket['__matches__'][0] = str(parts[0])
except ValueError:
pass
else:
# Return 1
return router.dynamic_routes[('<__dynamic__:str>',)][0], basket
raise NotFoundFYI: If you are on Python 3.9, you can see a representation of the source after compilation at router.find_route_src_compiled
Therefore, in general implementation requires you to:
- Define a router with a
getmethod; - Add one or more routes;
- Finalize the router (
router.finalize()); and - Call the router's
getmethod.
NOTE: You can call router.finalize(False) if you do not want to compile the source code into executable form. This is useful if you only intend to review the generated output.
Every time you call router.add you create one (1) new Route instance. Even if that one route is created with multiple methods, it generates a single instance. If you add() another Route that has a similar path structure (but, perhaps has differen methods) they will be grouped together into a RouteGroup. It is worth also noting that a RouteGroup is created the first time you call add(), but subsequent similar routes will reuse the existing grouping instance.
When you call finalize(), it is taking the defined route groups and arranging them into "nodes" in a hierarchical tree. A single node is a path segment. A Node instance can have one or more RouteGroup on it where the Node is the termination point for that path.
Perhaps an example is easier:
router.add("/path/to/<foo>", lambda: ...)
router.add("/path/to/<foo:int>", lambda: ...)
router.add("/path/to/different/<foo>", lambda: ...)
router.add("/path/to/different/<foo>", lambda: ..., methods=["one", "two"])The generated RouteGroup instances (3):
<RouteGroup: path=path/to/<foo:str> len=1>
<RouteGroup: path=path/to/<foo:int> len=1>
<RouteGroup: path=path/to/different/<foo:str> len=2>
The generated Route instances (4):
<Route: path=path/to/<foo:str>>
<Route: path=path/to/<foo:int>>
<Route: path=path/to/different/<foo:str>>
<Route: path=path/to/different/<foo:str>>
The Node Tree:
<Node: level=0>
<Node: part=path, level=1>
<Node: part=to, level=2>
<Node: part=different, level=3>
<Node: part=__dynamic__:str, level=4, groups=[<RouteGroup: path=path/to/different/<foo:str> len=2>], dynamic=True>
<Node: part=__dynamic__:int, level=3, groups=[<RouteGroup: path=path/to/<foo:int> len=1>], dynamic=True>
<Node: part=__dynamic__:str, level=3, groups=[<RouteGroup: path=path/to/<foo:str> len=1>], dynamic=True>
And, the generated source code:
def find_route(path, method, router, basket, extra):
parts = tuple(path[1:].split(router.delimiter))
num = len(parts)
# node=1 // part=path
if num > 1: # CHECK 1
if parts[0] == "path": # CHECK 4
# node=1.1 // part=to
if num > 2: # CHECK 1
if parts[1] == "to": # CHECK 4
# node=1.1.1 // part=different
if num > 3: # CHECK 1
if parts[2] == "different": # CHECK 4
# node=1.1.1.1 // part=__dynamic__:str
if num == 4: # CHECK 1
try:
basket['__matches__'][3] = str(parts[3])
except ValueError:
pass
else:
if method in frozenset({'one', 'two'}):
route_idx = 0
elif method in frozenset({'BASE'}):
route_idx = 1
else:
raise NoMethod
# Return 1.1.1.1
return router.dynamic_routes[('path', 'to', 'different', '<__dynamic__:str>')][route_idx], basket
# node=1.1.2 // part=__dynamic__:int
if num >= 3: # CHECK 1
try:
basket['__matches__'][2] = int(parts[2])
except ValueError:
pass
else:
if num == 3: # CHECK 5
# Return 1.1.2
return router.dynamic_routes[('path', 'to', '<__dynamic__:int>')][0], basket
# node=1.1.3 // part=__dynamic__:str
if num >= 3: # CHECK 1
try:
basket['__matches__'][2] = str(parts[2])
except ValueError:
pass
else:
if num == 3: # CHECK 5
# Return 1.1.3
return router.dynamic_routes[('path', 'to', '<__dynamic__:str>')][0], basket
raise NotFoundThe above example only shows routes that have a dynamic path segment in them (example: <foo>). But, there are other use cases that are covered differently:
- fully static paths - These are paths with no parameters (example:
/user/login). These are basically matched against a key/value store. - regex paths - If a route as a single regular expression match, then the whole route will be matched via regex. In general, this happens inline not too dissimilar than what we see in the above example.
- special regex paths - The router comes with a special
pathtype (example:<foo:path>) that can match on an expanded delimiter. This is also true for any regex that uses the path delimiter in it. These cannot be matched in the normal course since they are of unknown length.