Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add a dependency graph tool. #1132

Merged
merged 2 commits into from
Dec 14, 2020
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 2 additions & 1 deletion pex/tools/commands/__init__.py
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@
# Licensed under the Apache License, Version 2.0 (see LICENSE).

from pex.tools.command import Command
from pex.tools.commands.graph import Graph
from pex.tools.commands.info import Info
from pex.tools.commands.interpreter import Interpreter
from pex.tools.commands.venv import Venv
Expand All @@ -13,4 +14,4 @@

def all_commands():
# type: () -> Iterable[Command]
return Info(), Interpreter(), Venv()
return Info(), Interpreter(), Graph(), Venv()
116 changes: 116 additions & 0 deletions pex/tools/commands/digraph.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,116 @@
# Copyright 2020 Pants project contributors (see CONTRIBUTORS.md).
# Licensed under the Apache License, Version 2.0 (see LICENSE).

from __future__ import absolute_import

from pex.typing import TYPE_CHECKING

if TYPE_CHECKING:
from typing import Dict, IO, List, Mapping, Optional, Tuple

Value = Optional[str]
Attributes = Mapping[str, Value]


class DiGraph(object):
"""Renders a dot digraph built up from nodes and edges."""

@staticmethod
def _render_ID(value):
# type: (str) -> str
# See https://graphviz.org/doc/info/lang.html for the various forms of `ID`.
return '"{}"'.format(value.replace('"', '\\"'))

@classmethod
def _render_a_list(cls, attributes):
# type: (Attributes) -> str
# See https://graphviz.org/doc/info/lang.html for the `a_list` production.
return ", ".join(
"{name}={value}".format(name=name, value=cls._render_ID(value))
for name, value in attributes.items()
if value is not None
)

def __init__(
self,
name, # type: str
strict=True, # type: bool
**attributes # type: Value
):
# type: (...) -> None
"""
:param name: A name for the graph.
:param strict: Whether or not duplicate edges are collapsed into one edge.
"""
self._name = name
self._strict = strict
self._attributes = attributes # type: Attributes
self._nodes = {} # type: Dict[str, Attributes]
self._edges = [] # type: List[Tuple[str, str, Attributes]]

@property
def name(self):
return self._name

def add_node(
self,
name, # type: str
**attributes # type: Value
):
# type: (...) -> None
"""Adds a node to the graph.

This is done implicitly by add_edge for the nodes the edge connects, but may be useful when
the node is either isolated or else needs to be decorated with attributes.

:param name: The name of the node.
"""
self._nodes[name] = attributes

def add_edge(
self,
start, # type: str
end, # type: str
**attributes # type: Value
):
# type: (...) -> None
"""

:param start: The name of the start node.
:param end: The name of the end node.
:param attributes: Any extra attributes for the edge connecting the start node to the end
node.
"""
self._edges.append((start, end, attributes))

def emit(self, out):
# type: (IO[str]) -> None
"""Render the current state of this digraph to the given `out` stream.

:param out: A stream to render this digraph to. N/B.: Will not be flushed or closed.
"""

def emit_attr_stmt(
stmt, # type: str
attributes, # type: Attributes
):
# type: (...) -> None
# See https://graphviz.org/doc/info/lang.html for the `attr_stmt` production.
out.write(
"{statement} [{a_list}];\n".format(
statement=stmt, a_list=self._render_a_list(attributes)
)
)

if self._strict:
out.write("strict ")
out.write("digraph {name} {{\n".format(name=self._render_ID(self._name)))
emit_attr_stmt("graph", self._attributes)
for node, attributes in self._nodes.items():
emit_attr_stmt(self._render_ID(node), attributes)
for start, end, attributes in self._edges:
emit_attr_stmt(
"{start} -> {end}".format(start=self._render_ID(start), end=self._render_ID(end)),
attributes,
)
out.write("}\n")
170 changes: 170 additions & 0 deletions pex/tools/commands/graph.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,170 @@
# Copyright 2020 Pants project contributors (see CONTRIBUTORS.md).
# Licensed under the Apache License, Version 2.0 (see LICENSE).

from __future__ import absolute_import

import logging
import os
import tempfile
import threading
from argparse import ArgumentParser, Namespace
from contextlib import contextmanager

from pex.common import safe_mkdir
from pex.dist_metadata import requires_dists
from pex.interpreter import PythonInterpreter
from pex.pex import PEX
from pex.tools.command import Command, Ok, OutputMixin, Result, try_open_file, try_run_program
from pex.tools.commands.digraph import DiGraph
from pex.typing import TYPE_CHECKING
from pex.variables import ENV

if TYPE_CHECKING:
from typing import Iterator, IO, Tuple

logger = logging.getLogger(__name__)


class Graph(OutputMixin, Command):
"""Generates a dot graph of the dependencies contained in a PEX file."""
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So, generating a graph is definitely useful, but if this were to emit JSON, someone could fairly easily convert to DOT if they wanted to, and JSON is easier to parse if you want to do something other than render it.

Is there already a tool to dump the dependencies in a parseable format? That might be a higher priority.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It would also be nice to have a standard format that Pants can render in various ways, so that it's uniform across languages (when that's a thing).

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There is no such tool. I'm happy to circle back and add support for dumping some json format instead of DOT, but DOT was immediately a priority for me working on the resolve-from-pexes project.

Copy link
Member Author

@jsirois jsirois Dec 14, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'll point out DOT is that standard format https://graphviz.org/doc/info/lang.html - there are just not alot of parsers pre-written for it. There is one for Python though https://pypi.org/project/pydot/. All that said, I'll re-iterate I'm happy to circle back when Pants identifies a suitable alternate standard.

Copy link

@stuhood stuhood Dec 14, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm happy to circle back and add support for dumping some json format instead of DOT, but DOT was immediately a priority for me working on the resolve-from-pexes project. [...] I'll point out DOT is that standard format https://graphviz.org/doc/info/lang.html - there are just not alot of parsers pre-written for it.

If the key driver is to be able to parse it, and being able to visualize it is secondary, it would seemingly take less code to dump an adjacency list in JSON (no standard necessary, I don't think).

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The pydot project does the parsing of DOT format and provides a Graph with get_edges, etc. As to an adjacency list in json being simple - agreed. So today I do that here in PEX and tomorrow you need this from coursier. Will it produce the same format? Do we need to hack on it to do so? Looking into some standard and then reporting back to the PEX project to influence its choice seems wise instead of ginning up something ad-hoc. At least DOT has a spec.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

And to highlight again the above - the key driver for me (and Pex) here was to look at the graph to help sort through the resolve-from-pex project. Pants may have other needs and that's fine - proposing a standard with deliberation would be great if that's what Pants wants to lobby for.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I do not think that you need a standard here (and a spec as large as DOT's is arguably counterproductive: how much of the DOT spec would you have to implement to conform?) Each set of @rules that wraps a resolver would just need to emit an adjacency list parsed from whatever format was natural for those rules... you would not need an abstract file parser, just an in-memory data type for the adjacency list.

As to an adjacency list in json being simple - agreed. So today I do that here in PEX and tomorrow you need this from coursier.

It already supports JSON output: https://get-coursier.io/docs/json-report

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

To clarify, I think it's fine to merge this as-is and revisit as needed.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Here - I filed the initial feature request for you: #1137 It needs some filling out though with the actual requirements Pants has unless Pants truly just wants an adjacency list with no extra data.


@staticmethod
def _create_dependency_graph(pex):
# type: (PEX) -> DiGraph
graph = DiGraph(
pex.path(),
fontsize="14",
labelloc="t",
label="Dependency graph of {} for interpreter {} ({})".format(
pex.path(), pex.interpreter.binary, pex.interpreter.identity.requirement
),
)
marker_environment = pex.interpreter.identity.env_markers.copy()
marker_environment["extra"] = []
present_dists = frozenset(dist.project_name for dist in pex.activate())
for dist in pex.activate():
graph.add_node(
name=dist.project_name,
label="{name} {version}".format(name=dist.project_name, version=dist.version),
URL="https://pypi.org/project/{name}/{version}".format(
name=dist.project_name, version=dist.version
),
target="_blank",
)
for req in requires_dists(dist):
if (
req.project_name not in present_dists
and req.marker
and not req.marker.evaluate(environment=marker_environment)
):
graph.add_node(
name=req.project_name,
color="lightgrey",
style="filled",
tooltip="inactive requirement",
URL="https://pypi.org/project/{name}".format(name=req.project_name),
target="_blank",
)
graph.add_edge(
start=dist.project_name,
end=req.project_name,
label=str(req) if (req.specifier or req.marker) else None,
fontsize="10",
)
return graph

def add_arguments(self, parser):
# type: (ArgumentParser) -> None
self.add_output_option(parser, entity="dot graph")
parser.add_argument(
"-r",
"--render",
action="store_true",
help="Attempt to render the graph.",
)
parser.add_argument(
"-f",
"--format",
default="svg",
help="The format to render the graph in.",
)
parser.add_argument(
"--open",
action="store_true",
help="Attempt to open the graph in the system viewer (implies --render).",
)

@staticmethod
def _dot(
options, # type: Namespace
graph, # type: DiGraph
render_fp, # type: IO
):
# type: (...) -> Result
read_fd, write_fd = os.pipe()

def emit():
with os.fdopen(write_fd, "w") as fp:
graph.emit(fp)

emit_thread = threading.Thread(name="{} Emitter".format(__name__), target=emit)
emit_thread.daemon = True
emit_thread.start()

try:
return try_run_program(
"dot",
url="https://graphviz.org/",
error="Failed to render dependency graph for {}.".format(graph.name),
args=["-T", options.format],
stdin=read_fd,
stdout=render_fp,
)
finally:
emit_thread.join()

@contextmanager
def _output_for_open(self, options):
# type: (Namespace) -> Iterator[Tuple[IO, str]]
if self.is_stdout(options):
tmpdir = os.path.join(ENV.PEX_ROOT, "tmp")
safe_mkdir(tmpdir)
with tempfile.NamedTemporaryFile(
prefix="{}.".format(__name__),
suffix=".deps.{}".format(options.format),
dir=tmpdir,
delete=False,
) as tmp_out:
yield tmp_out, tmp_out.name
return

with self.output(options, binary=True) as out:
yield out, out.name

def run(
self,
pex, # type: PEX
options, # type: Namespace
):
# type: (...) -> Result
graph = self._create_dependency_graph(pex)
if not (options.render or options.open):
with self.output(options) as out:
graph.emit(out)
return Ok()

if not options.open:
with self.output(options, binary=True) as out:
return self._dot(options, graph, out)

with self._output_for_open(options) as (out, open_path):
result = self._dot(options, graph, out)
if result.is_error:
return result

return try_open_file(
open_path,
error="Failed to open dependency graph of {} rendered in {} for viewing.".format(
pex.path(), open_path
),
)