( )⚙ D2834 wireproto: support /api/* URL space for exposing APIs

This is an archive of the discontinued Mercurial Phabricator instance.

Differential D2834

wireproto: support /api/* URL space for exposing APIs
ClosedPublic

Authored by indygreg on Mar 12 2018, 8:07 PM.

Details

Reviewers
durin42
Group Reviewers
hg-reviewers
Commits
rHG1cfef5693203: wireproto: support /api/* URL space for exposing APIs
Summary

I will soon be introducing a new version of the HTTP wire protocol.
One of the things I want to change with it is the URL routing.
I want to rely on URL paths to define endpoints rather than the
"cmd" query string argument. That should be pretty straightforward.

I was thinking about what URL space to reserve for the new protocol.
We /could/ put everything at a top-level path. e.g.
/wireproto/* or /http-v2-wireproto/*. However, these constrain us
a bit because they assume there will only be 1 API: version 2 of
the HTTP wire protocol. I think there is room to grow multiple
APIs. For example, there may someday be a proper JSON API to query
or even manipulate the repository. And I don't think we should have
to create a new top-level URL space for each API nor should we
attempt to shoehorn each future API into the same shared URL space:
that would just be too chaotic.

This commits reserves the /api/* URL space for all our future API
needs. Essentially, all requests to /api/* get routed to a new WSGI
handler. By default, it 404's the entire URL space unless the
"api server" feature is enabled. When enabled, requests to "/api"
list available APIs. URLs of the form /api/<name>/* are reserved for
a particular named API. Behavior within each API is left up to that
API. So, we can grow new APIs easily without worrying about URL
space conflicts.

APIs can be registered by adding entries to a global dict. This allows
extensions to provide their own APIs should they choose to do so.
This is probably a premature feature. But IMO the code is easier
to read if we're not dealing with API-specific behavior like config
option querying inline.

To prove it works, we implement a very basic API for version 2
of the HTTP wire protocol. It does nothing of value except
facilitate testing of the /api/* URL space.

We currently emit plain text responses for all /api/* endpoints.
There's definitely room to look at Accept and other request headers
to vary the response format. But we have to start somewhere.

Diff Detail

Repository
rHG Mercurial
Lint
Lint Skipped
Unit
Unit Tests Skipped

Event Timeline

indygreg created this revision.Mar 12 2018, 8:07 PM
mharbison72 added a subscriber: mharbison72.Mar 13 2018, 9:56 AM

Will this cause conflicts with hg serve -S or hgwebdir with subrepo/virtual paths that start with 'api'? (This was the reason I used the LFS URI starting with '.hg/', knowing that isn't committable.)

indygreg added a comment.Mar 13 2018, 11:46 AM
In D2834#45666, @mharbison72 wrote:

Will this cause conflicts with hg serve -S or hgwebdir with subrepo/virtual paths that start with 'api'? (This was the reason I used the LFS URI starting with '.hg/', knowing that isn't committable.)

Bleh.

So what I'm hearing is that because of hgwebdir's virtual paths and support for serving subrepos, pretty much the entire URL space is dangerous because it can conflict with the path of a virtual repo or subrepo. I suppose we already have problems with existing URLs, like /rev/.

I agree that putting things under .hg/ in the URL space seems to be a viable workaround.

Changing the URL prefix isn't a big deal as far as the code is concerned. We can make that change at any time since we have no BC guarantees at this juncture.

mharbison72 added a comment.Mar 13 2018, 1:36 PM
In D2834#45696, @indygreg wrote:
In D2834#45666, @mharbison72 wrote:

Will this cause conflicts with hg serve -S or hgwebdir with subrepo/virtual paths that start with 'api'? (This was the reason I used the LFS URI starting with '.hg/', knowing that isn't committable.)

Bleh.
So what I'm hearing is that because of hgwebdir's virtual paths and support for serving subrepos, pretty much the entire URL space is dangerous because it can conflict with the path of a virtual repo or subrepo. I suppose we already have problems with existing URLs, like /rev/.

I think so. I remember when Yuya suggested putting the /index virtual file into hgwebdir a year ago, I had to dance around the fact that 'index' could have been a repo being served up. I'm guessing /rev/ et al. is less of a problem, because presumably they've existed since the beginning?

I agree that putting things under .hg/ in the URL space seems to be a viable workaround.
Changing the URL prefix isn't a big deal as far as the code is concerned. We can make that change at any time since we have no BC guarantees at this juncture.

I'm fine with landing this as-is and changing it later if needed. I haven't checked this myself, I was just going on a vague memory of issues that I ran into, and wondered if you had considered it.

indygreg updated this revision to Diff 6998.Mar 13 2018, 3:11 PM
indygreg updated this revision to Diff 7009.Mar 13 2018, 9:16 PM
indygreg added a comment.Mar 19 2018, 7:37 PM

So, assuming we can't introduce new top-level URLs without risking conflicts with subrepos or hgwebdir virtual repos, for at least the wire protocol, assuming we keep the ?cmd=capabilities mechanism for the initial server handshake, we can add something that defines where API requests should be routed. We can then allow server operators to reconfigure the base URL for these requests.

durin42 accepted this revision.Mar 21 2018, 3:18 PM
This revision is now accepted and ready to land.Mar 21 2018, 3:18 PM
indygreg updated this revision to Diff 7194.Mar 21 2018, 6:19 PM
durin42 accepted this revision.Mar 21 2018, 6:21 PM
Closed by commit rHG1cfef5693203: wireproto: support /api/* URL space for exposing APIs (authored by indygreg). · Explain WhyMar 21 2018, 8:57 PM
This revision was automatically updated to reflect the committed changes.
yuja added a subscriber: yuja.Mar 23 2018, 11:22 AM

Can we expect the BC of hgwebdir will be fixed before the release?

I like the /api/ namespace, but breaking existing hgweb instances
wouldn't be acceptable.

indygreg added a comment.Mar 23 2018, 2:20 PM
In D2834#47334, @yuja wrote:

Can we expect the BC of hgwebdir will be fixed before the release?
I like the /api/ namespace, but breaking existing hgweb instances
wouldn't be acceptable.

D2936

Revision Contents
Changeset List

Diff 6979

mercurial/configitems.py

default=False, default=False,
) )
coreconfigitem('experimental', 'update.atomic-file', coreconfigitem('experimental', 'update.atomic-file',
default=False, default=False,
) )
coreconfigitem('experimental', 'sshpeer.advertise-v2', coreconfigitem('experimental', 'sshpeer.advertise-v2',
default=False, default=False,
) )
coreconfigitem('experimental', 'web.apiserver',
default=False,
)
coreconfigitem('experimental', 'web.api.http-v2',
default=False,
)
coreconfigitem('experimental', 'xdiff', coreconfigitem('experimental', 'xdiff',
default=False, default=False,
) )
coreconfigitem('extensions', '.*', coreconfigitem('extensions', '.*',
default=None, default=None,
generic=True, generic=True,
) )
coreconfigitem('extdata', '.*', coreconfigitem('extdata', '.*',

mercurial/hgweb/hgweb_mod.py

encoding.encoding = rctx.config('web', 'encoding') encoding.encoding = rctx.config('web', 'encoding')
rctx.repo.ui.environ = req.rawenv rctx.repo.ui.environ = req.rawenv
if rctx.csp: if rctx.csp:
# hgwebdir may have added CSP header. Since we generate our own, # hgwebdir may have added CSP header. Since we generate our own,
# replace it. # replace it.
res.headers['Content-Security-Policy'] = rctx.csp res.headers['Content-Security-Policy'] = rctx.csp
# /api/* is reserved for various API implementations. Dispatch
# accordingly.
if req.dispatchparts and req.dispatchparts[0] == b'api':
wireprotoserver.handlewsgiapirequest(rctx, req, res,
self.check_perm)
return res.sendresponse()
handled = wireprotoserver.handlewsgirequest( handled = wireprotoserver.handlewsgirequest(
rctx, req, res, self.check_perm) rctx, req, res, self.check_perm)
if handled: if handled:
return res.sendresponse() return res.sendresponse()
# Old implementations of hgweb supported dispatching the request via # Old implementations of hgweb supported dispatching the request via
# the initial query string parameter instead of using PATH_INFO. # the initial query string parameter instead of using PATH_INFO.
# If PATH_INFO is present (signaled by ``req.dispatchpath`` having # If PATH_INFO is present (signaled by ``req.dispatchpath`` having

mercurial/wireprotoserver.py

urlreq = util.urlreq urlreq = util.urlreq
HTTP_OK = 200 HTTP_OK = 200
HGTYPE = 'application/mercurial-0.1' HGTYPE = 'application/mercurial-0.1'
HGTYPE2 = 'application/mercurial-0.2' HGTYPE2 = 'application/mercurial-0.2'
HGERRTYPE = 'application/hg-error' HGERRTYPE = 'application/hg-error'
HTTPV2 = wireprototypes.HTTPV2
SSHV1 = wireprototypes.SSHV1 SSHV1 = wireprototypes.SSHV1
SSHV2 = wireprototypes.SSHV2 SSHV2 = wireprototypes.SSHV2
def decodevaluefromheaders(req, headerprefix): def decodevaluefromheaders(req, headerprefix):
"""Decode a long value from multiple HTTP request headers. """Decode a long value from multiple HTTP request headers.
Returns the value as a bytes, not a str. Returns the value as a bytes, not a str.
""" """
res.headers[k] = v res.headers[k] = v
res.status = hgwebcommon.statusmessage(e.code, pycompat.bytestr(e)) res.status = hgwebcommon.statusmessage(e.code, pycompat.bytestr(e))
# TODO This response body assumes the failed command was # TODO This response body assumes the failed command was
# "unbundle." That assumption is not always valid. # "unbundle." That assumption is not always valid.
res.setbodybytes('0\n%s\n' % pycompat.bytestr(e)) res.setbodybytes('0\n%s\n' % pycompat.bytestr(e))
return True return True
def handlewsgiapirequest(rctx, req, res, checkperm):
"""Handle requests to /api/*."""
assert req.dispatchparts[0] == b'api'
repo = rctx.repo
# This whole URL space is experimental for now. But we want to
# reserve the URL space. So, 404 all URLs if the feature isn't enabled.
if not repo.ui.configbool('experimental', 'web.apiserver'):
res.status = b'404 Not Found'
res.headers[b'Content-Type'] = b'text/plain'
res.setbodybytes(_('Experimental API server endpoint not enabled'))
return
# The URL space is /api/<protocol>/*. The structure of URLs under varies
# by <protocol>.
# Registered APIs are made available via config options of the name of
# the protocol.
availableapis = set()
for k, v in API_HANDLERS.items():
section, option = v['config']
if repo.ui.configbool(section, option):
availableapis.add(k)
# Requests to /api/ list available APIs.
if req.dispatchparts == [b'api']:
res.status = b'200 OK'
res.headers[b'Content-Type'] = b'text/plain'
lines = [_('APIs can be accessed at /api/<name>, where <name> can be '
'one of the following:\n')]
if availableapis:
lines.extend(sorted(availableapis))
else:
lines.append(_('(no available APIs)\n'))
res.setbodybytes(b'\n'.join(lines))
return
proto = req.dispatchparts[1]
if proto not in API_HANDLERS:
res.status = b'404 Not Found'
res.headers[b'Content-Type'] = b'text/plain'
res.setbodybytes(_('Unknown API: %s\nKnown APIs: %s') % (
proto, b', '.join(sorted(availableapis))))
return
if proto not in availableapis:
res.status = b'404 Not Found'
res.headers[b'Content-Type'] = b'text/plain'
res.setbodybytes(_('API %s not enabled\n') % proto)
return
API_HANDLERS[proto]['handler'](rctx, req, res, checkperm,
req.dispatchparts[2:])
def _handlehttpv2request(rctx, req, res, checkperm, urlparts):
res.status = b'200 OK'
res.headers[b'Content-Type'] = b'text/plain'
res.setbodybytes(b'/'.join(urlparts) + b'/')
# Maps API name to metadata so custom API can be registered.
API_HANDLERS = {
HTTPV2: {
'config': ('experimental', 'web.api.http-v2'),
'handler': _handlehttpv2request,
},
}
def _httpresponsetype(ui, req, prefer_uncompressed): def _httpresponsetype(ui, req, prefer_uncompressed):
"""Determine the appropriate response type and compression settings. """Determine the appropriate response type and compression settings.
Returns a tuple of (mediatype, compengine, engineopts). Returns a tuple of (mediatype, compengine, engineopts).
""" """
# Determine the response media type and compression engine based # Determine the response media type and compression engine based
# on the request parameters. # on the request parameters.
protocaps = decodevaluefromheaders(req, 'X-HgProto').split(' ') protocaps = decodevaluefromheaders(req, 'X-HgProto').split(' ')

mercurial/wireprototypes.py

# Copyright 2018 Gregory Szorc <gregory.szorc@gmail.com> # Copyright 2018 Gregory Szorc <gregory.szorc@gmail.com>
# #
# This software may be used and distributed according to the terms of the # This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version. # GNU General Public License version 2 or any later version.
from __future__ import absolute_import from __future__ import absolute_import
import abc import abc
# Names of the SSH protocol implementations. # Names of the SSH protocol implementations.
SSHV1 = 'ssh-v1' SSHV1 = 'ssh-v1'
# This is advertised over the wire. Incremental the counter at the end # These are advertised over the wire. Increment the counters at the end
# to reflect BC breakages. # to reflect BC breakages.
SSHV2 = 'exp-ssh-v2-0001' SSHV2 = 'exp-ssh-v2-0001'
HTTPV2 = 'exp-http-v2-0001'
# All available wire protocol transports. # All available wire protocol transports.
TRANSPORTS = { TRANSPORTS = {
SSHV1: { SSHV1: {
'transport': 'ssh', 'transport': 'ssh',
'version': 1, 'version': 1,
}, },
SSHV2: { SSHV2: {
'transport': 'ssh', 'transport': 'ssh',
'version': 2, 'version': 2,
}, },
'http-v1': { 'http-v1': {
'transport': 'http', 'transport': 'http',
'version': 1, 'version': 1,
},
HTTPV2: {
'transport': 'http',
'version': 2,
} }
} }
class bytesresponse(object): class bytesresponse(object):
"""A wire protocol response consisting of raw bytes.""" """A wire protocol response consisting of raw bytes."""
def __init__(self, data): def __init__(self, data):
self.data = data self.data = data

tests/test-http-api-httpv2.t

  • This file was added.
$ hg init server
$ cat > server/.hg/hgrc << EOF
> [experimental]
> web.apiserver = true
> EOF
$ hg -R server serve -p $HGPORT -d --pid-file hg.pid
$ cat hg.pid > $DAEMON_PIDS
HTTP v2 protocol not enabled by default
$ get-with-headers.py $LOCALIP:$HGPORT api/exp-http-v2-0001 -
404 Not Found
content-length: 33
content-type: text/plain
API exp-http-v2-0001 not enabled
[1]
Restart server with support for HTTP v2 API
$ killdaemons.py
$ cat > server/.hg/hgrc << EOF
> [experimental]
> web.apiserver = true
> web.api.http-v2 = true
> EOF
$ hg -R server serve -p $HGPORT -d --pid-file hg.pid
$ cat hg.pid > $DAEMON_PIDS
Requests simply echo their path (for now)
$ get-with-headers.py $LOCALIP:$HGPORT api/exp-http-v2-0001/path1/path2 -
200 OK
content-length: 12
content-type: text/plain
path1/path2/ (no-eol)

tests/test-http-api.t

  • This file was added.
$ hg init server
$ hg -R server serve -p $HGPORT -d --pid-file hg.pid
$ cat hg.pid > $DAEMON_PIDS
Request to /api fails unless web.apiserver is enabled
$ get-with-headers.py $LOCALIP:$HGPORT api -
404 Not Found
content-length: 44
content-type: text/plain
Experimental API server endpoint not enabled (no-eol)
[1]
$ get-with-headers.py $LOCALIP:$HGPORT api/ -
404 Not Found
content-length: 44
content-type: text/plain
Experimental API server endpoint not enabled (no-eol)
[1]
Restart server with support for API server
$ killdaemons.py
$ cat > server/.hg/hgrc << EOF
> [experimental]
> web.apiserver = true
> EOF
$ hg -R server serve -p $HGPORT -d --pid-file hg.pid
$ cat hg.pid > $DAEMON_PIDS
/api lists available APIs (empty since none are available by default)
$ get-with-headers.py $LOCALIP:$HGPORT api -
200 OK
content-length: 100
content-type: text/plain
APIs can be accessed at /api/<name>, where <name> can be one of the following:
(no available APIs)
$ get-with-headers.py $LOCALIP:$HGPORT api/ -
200 OK
content-length: 100
content-type: text/plain
APIs can be accessed at /api/<name>, where <name> can be one of the following:
(no available APIs)
Accessing an unknown API yields a 404
$ get-with-headers.py $LOCALIP:$HGPORT api/unknown -
404 Not Found
content-length: 33
content-type: text/plain
Unknown API: unknown
Known APIs: (no-eol)
[1]
Accessing a known but not enabled API yields a different error
$ get-with-headers.py $LOCALIP:$HGPORT api/exp-http-v2-0001 -
404 Not Found
content-length: 33
content-type: text/plain
API exp-http-v2-0001 not enabled
[1]
Restart server with support for HTTP v2 API
$ killdaemons.py
$ cat > server/.hg/hgrc << EOF
> [experimental]
> web.apiserver = true
> web.api.http-v2 = true
> EOF
$ hg -R server serve -p $HGPORT -d --pid-file hg.pid
$ cat hg.pid > $DAEMON_PIDS
/api lists the HTTP v2 protocol as available
$ get-with-headers.py $LOCALIP:$HGPORT api -
200 OK
content-length: 96
content-type: text/plain
APIs can be accessed at /api/<name>, where <name> can be one of the following:
exp-http-v2-0001 (no-eol)
$ get-with-headers.py $LOCALIP:$HGPORT api/ -
200 OK
content-length: 96
content-type: text/plain
APIs can be accessed at /api/<name>, where <name> can be one of the following:
exp-http-v2-0001 (no-eol)