pyrogram/compiler/api/compiler.py

603 lines
20 KiB
Python

# Pyrogram - Telegram MTProto API Client Library for Python
# Copyright (C) 2017-present Dan <https://github.com/delivrance>
#
# This file is part of Pyrogram.
#
# Pyrogram is free software: you can redistribute it and/or modify
# it under the terms of the GNU Lesser General Public License as published
# by the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# Pyrogram is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU Lesser General Public License for more details.
#
# You should have received a copy of the GNU Lesser General Public License
# along with Pyrogram. If not, see <http://www.gnu.org/licenses/>.
import os
import re
import shutil
from functools import partial
from pathlib import Path
from typing import NamedTuple, List, Tuple
# from autoflake import fix_code
# from black import format_str, FileMode
HOME_PATH = Path("compiler/api")
DESTINATION_PATH = Path("pyrogram/raw")
NOTICE_PATH = "NOTICE"
SECTION_RE = re.compile(r"---(\w+)---")
LAYER_RE = re.compile(r"//\sLAYER\s(\d+)")
COMBINATOR_RE = re.compile(r"^([\w.]+)#([0-9a-f]+)\s(?:.*)=\s([\w<>.]+);$", re.MULTILINE)
ARGS_RE = re.compile(r"[^{](\w+):([\w?!.<>#]+)")
FLAGS_RE = re.compile(r"flags(\d?)\.(\d+)\?")
FLAGS_RE_2 = re.compile(r"flags(\d?)\.(\d+)\?([\w<>.]+)")
FLAGS_RE_3 = re.compile(r"flags(\d?):#")
INT_RE = re.compile(r"int(\d+)")
CORE_TYPES = ["int", "long", "int128", "int256", "double", "bytes", "string", "Bool", "true"]
WARNING = """
# # # # # # # # # # # # # # # # # # # # # # # #
# !!! WARNING !!! #
# This is a generated file! #
# All changes made in this file will be lost! #
# # # # # # # # # # # # # # # # # # # # # # # #
""".strip()
# noinspection PyShadowingBuiltins
open = partial(open, encoding="utf-8")
types_to_constructors = {}
types_to_functions = {}
constructors_to_functions = {}
namespaces_to_types = {}
namespaces_to_constructors = {}
namespaces_to_functions = {}
class Combinator(NamedTuple):
section: str
qualname: str
namespace: str
name: str
id: str
has_flags: bool
args: List[Tuple[str, str]]
qualtype: str
typespace: str
type: str
def snake(s: str):
# https://stackoverflow.com/q/1175208
s = re.sub(r"(.)([A-Z][a-z]+)", r"\1_\2", s)
return re.sub(r"([a-z0-9])([A-Z])", r"\1_\2", s).lower()
def camel(s: str):
return "".join([i[0].upper() + i[1:] for i in s.split("_")])
# noinspection PyShadowingBuiltins, PyShadowingNames
def get_type_hint(type: str) -> str:
is_flag = FLAGS_RE.match(type)
is_core = False
if is_flag:
type = type.split("?")[1]
if type in CORE_TYPES:
is_core = True
if type == "long" or "int" in type:
type = "int"
elif type == "double":
type = "float"
elif type == "string":
type = "str"
elif type in ["Bool", "true"]:
type = "bool"
else: # bytes and object
type = "bytes"
if type in ["Object", "!X"]:
return "TLObject"
if re.match("^vector", type, re.I):
is_core = True
sub_type = type.split("<")[1][:-1]
type = f"List[{get_type_hint(sub_type)}]"
if is_core:
return f"Optional[{type}] = None" if is_flag else type
else:
ns, name = type.split(".") if "." in type else ("", type)
type = f'"raw.base.' + ".".join([ns, name]).strip(".") + '"'
return f'{type}{" = None" if is_flag else ""}'
def sort_args(args):
"""Put flags at the end"""
args = args.copy()
flags = [i for i in args if FLAGS_RE.match(i[1])]
for i in flags:
args.remove(i)
for i in args[:]:
if re.match(r"flags\d?", i[0]) and i[1] == "#":
args.remove(i)
return args + flags
def remove_whitespaces(source: str) -> str:
"""Remove whitespaces from blank lines"""
lines = source.split("\n")
for i, _ in enumerate(lines):
if re.match(r"^\s+$", lines[i]):
lines[i] = ""
return "\n".join(lines)
def get_docstring_arg_type(t: str, is_list: bool = False, is_pyrogram_type: bool = False):
if t in CORE_TYPES:
if t == "long":
return "``int`` ``64-bit``"
elif "int" in t:
size = INT_RE.match(t)
return f"``int`` ``{size.group(1)}-bit``" if size else "``int`` ``32-bit``"
elif t == "double":
return "``float`` ``64-bit``"
elif t == "string":
return "``str``"
elif t == "true":
return "``bool``"
else:
return f"``{t.lower()}``"
elif t == "TLObject" or t == "X":
return "Any object from :obj:`~pyrogram.raw.types`"
elif t == "!X":
return "Any method from :obj:`~pyrogram.raw.functions`"
elif t.lower().startswith("vector"):
return "List of " + get_docstring_arg_type(t.split("<", 1)[1][:-1], True)
else:
return f":obj:`{t} <pyrogram.raw.base.{t}>`"
def get_references(t: str, kind: str):
if kind == "constructors":
t = constructors_to_functions.get(t)
elif kind == "types":
t = types_to_functions.get(t)
else:
raise ValueError("Invalid kind")
if t:
return "\n ".join(
f"- :obj:`{i} <pyrogram.raw.functions.{i}>`"
for i in t
), len(t)
return None, 0
# noinspection PyShadowingBuiltins
def start(format: bool = False):
shutil.rmtree(DESTINATION_PATH / "types", ignore_errors=True)
shutil.rmtree(DESTINATION_PATH / "functions", ignore_errors=True)
shutil.rmtree(DESTINATION_PATH / "base", ignore_errors=True)
with open(HOME_PATH / "source/auth_key.tl") as f1, \
open(HOME_PATH / "source/sys_msgs.tl") as f2, \
open(HOME_PATH / "source/main_api.tl") as f3:
schema = (f1.read() + f2.read() + f3.read()).splitlines()
with open(HOME_PATH / "template/type.txt") as f1, \
open(HOME_PATH / "template/combinator.txt") as f2:
type_tmpl = f1.read()
combinator_tmpl = f2.read()
with open(NOTICE_PATH, encoding="utf-8") as f:
notice = []
for line in f.readlines():
notice.append(f"# {line}".strip())
notice = "\n".join(notice)
section = None
layer = None
combinators = []
for line in schema:
# Check for section changer lines
section_match = SECTION_RE.match(line)
if section_match:
section = section_match.group(1)
continue
# Save the layer version
layer_match = LAYER_RE.match(line)
if layer_match:
layer = layer_match.group(1)
continue
combinator_match = COMBINATOR_RE.match(line)
if combinator_match:
# noinspection PyShadowingBuiltins
qualname, id, qualtype = combinator_match.groups()
namespace, name = qualname.split(".") if "." in qualname else ("", qualname)
name = camel(name)
qualname = ".".join([namespace, name]).lstrip(".")
typespace, type = qualtype.split(".") if "." in qualtype else ("", qualtype)
type = camel(type)
qualtype = ".".join([typespace, type]).lstrip(".")
# Pingu!
has_flags = not not FLAGS_RE_3.findall(line)
args = ARGS_RE.findall(line)
# Fix arg name being "self" (reserved python keyword)
for i, item in enumerate(args):
if item[0] == "self":
args[i] = ("is_self", item[1])
combinator = Combinator(
section=section,
qualname=qualname,
namespace=namespace,
name=name,
id=f"0x{id}",
has_flags=has_flags,
args=args,
qualtype=qualtype,
typespace=typespace,
type=type
)
combinators.append(combinator)
for c in combinators:
qualtype = c.qualtype
if qualtype.startswith("Vector"):
qualtype = qualtype.split("<")[1][:-1]
d = types_to_constructors if c.section == "types" else types_to_functions
if qualtype not in d:
d[qualtype] = []
d[qualtype].append(c.qualname)
if c.section == "types":
key = c.namespace
if key not in namespaces_to_types:
namespaces_to_types[key] = []
if c.type not in namespaces_to_types[key]:
namespaces_to_types[key].append(c.type)
for k, v in types_to_constructors.items():
for i in v:
try:
constructors_to_functions[i] = types_to_functions[k]
except KeyError:
pass
# import json
# print(json.dumps(namespaces_to_types, indent=2))
for qualtype in types_to_constructors:
typespace, type = qualtype.split(".") if "." in qualtype else ("", qualtype)
dir_path = DESTINATION_PATH / "base" / typespace
module = type
if module == "Updates":
module = "UpdatesT"
os.makedirs(dir_path, exist_ok=True)
constructors = sorted(types_to_constructors[qualtype])
constr_count = len(constructors)
items = "\n ".join([f"- :obj:`{c} <pyrogram.raw.types.{c}>`" for c in constructors])
docstring = f"This base type has {constr_count} constructor{'s' if constr_count > 1 else ''} available.\n\n"
docstring += f" Constructors:\n .. hlist::\n :columns: 2\n\n {items}"
references, ref_count = get_references(qualtype, "types")
if references:
docstring += f"\n\n See Also:\n This object can be returned by " \
f"{ref_count} method{'s' if ref_count > 1 else ''}:" \
f"\n\n .. hlist::\n :columns: 2\n\n " + references
with open(dir_path / f"{snake(module)}.py", "w") as f:
f.write(
type_tmpl.format(
notice=notice,
warning=WARNING,
docstring=docstring,
name=type,
qualname=qualtype,
types=", ".join([f"raw.types.{c}" for c in constructors]),
doc_name=snake(type).replace("_", "-")
)
)
for c in combinators:
sorted_args = sort_args(c.args)
arguments = (
(", *, " if c.args else "") +
(", ".join(
[f"{i[0]}: {get_type_hint(i[1])}"
for i in sorted_args]
) if sorted_args else "")
)
fields = "\n ".join(
[f"self.{i[0]} = {i[0]} # {i[1]}"
for i in sorted_args]
) if sorted_args else "pass"
docstring = ""
docstring_args = []
for i, arg in enumerate(sorted_args):
arg_name, arg_type = arg
is_optional = FLAGS_RE.match(arg_type)
flag_number = is_optional.group(1) if is_optional else -1
arg_type = arg_type.split("?")[-1]
docstring_args.append(
"{}{}: {}".format(
arg_name,
" (optional)".format(flag_number) if is_optional else "",
get_docstring_arg_type(arg_type, is_pyrogram_type=c.namespace == "pyrogram")
)
)
if c.section == "types":
docstring += f"This object is a constructor of the base type :obj:`~pyrogram.raw.base.{c.qualtype}`.\n\n"
else:
docstring += f"Telegram API method.\n\n"
docstring += f" Details:\n - Layer: ``{layer}``\n - ID: ``{c.id[2:].upper()}``\n\n"
if docstring_args:
docstring += " Parameters:\n " + "\n ".join(docstring_args)
else:
docstring += " **No parameters required.**"
if c.section == "functions":
docstring += "\n\n Returns:\n " + get_docstring_arg_type(c.qualtype)
else:
references, count = get_references(c.qualname, "constructors")
if references:
docstring += f"\n\n See Also:\n This object can be returned by " \
f"{count} method{'s' if count > 1 else ''}:" \
f"\n\n .. hlist::\n :columns: 2\n\n " + references
write_types = read_types = "" if c.has_flags else "# No flags\n "
for arg_name, arg_type in c.args:
flag = FLAGS_RE_2.match(arg_type)
if re.match(r"flags\d?", arg_name) and arg_type == "#":
write_flags = []
for i in c.args:
flag = FLAGS_RE_2.match(i[1])
if flag:
if arg_name != f"flags{flag.group(1)}":
continue
if flag.group(3) == "true" or flag.group(3).startswith("Vector"):
write_flags.append(f"{arg_name} |= (1 << {flag.group(2)}) if self.{i[0]} else 0")
else:
write_flags.append(
f"{arg_name} |= (1 << {flag.group(2)}) if self.{i[0]} is not None else 0")
write_flags = "\n ".join([
f"{arg_name} = 0",
"\n ".join(write_flags),
f"b.write(Int({arg_name}))\n "
])
write_types += write_flags
read_types += f"\n {arg_name} = Int.read(b)\n "
continue
if flag:
number, index, flag_type = flag.groups()
if flag_type == "true":
read_types += "\n "
read_types += f"{arg_name} = True if flags{number} & (1 << {index}) else False"
elif flag_type in CORE_TYPES:
write_types += "\n "
write_types += f"if self.{arg_name} is not None:\n "
write_types += f"b.write({flag_type.title()}(self.{arg_name}))\n "
read_types += "\n "
read_types += f"{arg_name} = {flag_type.title()}.read(b) if flags{number} & (1 << {index}) else None"
elif "vector" in flag_type.lower():
sub_type = arg_type.split("<")[1][:-1]
write_types += "\n "
write_types += f"if self.{arg_name}:\n "
write_types += "b.write(Vector(self.{}{}))\n ".format(
arg_name, f", {sub_type.title()}" if sub_type in CORE_TYPES else ""
)
read_types += "\n "
read_types += "{} = TLObject.read(b{}) if flags{} & (1 << {}) else []\n ".format(
arg_name, f", {sub_type.title()}" if sub_type in CORE_TYPES else "", number, index
)
else:
write_types += "\n "
write_types += f"if self.{arg_name} is not None:\n "
write_types += f"b.write(self.{arg_name}.write())\n "
read_types += "\n "
read_types += f"{arg_name} = TLObject.read(b) if flags{number} & (1 << {index}) else None\n "
else:
if arg_type in CORE_TYPES:
write_types += "\n "
write_types += f"b.write({arg_type.title()}(self.{arg_name}))\n "
read_types += "\n "
read_types += f"{arg_name} = {arg_type.title()}.read(b)\n "
elif "vector" in arg_type.lower():
sub_type = arg_type.split("<")[1][:-1]
write_types += "\n "
write_types += "b.write(Vector(self.{}{}))\n ".format(
arg_name, f", {sub_type.title()}" if sub_type in CORE_TYPES else ""
)
read_types += "\n "
read_types += "{} = TLObject.read(b{})\n ".format(
arg_name, f", {sub_type.title()}" if sub_type in CORE_TYPES else ""
)
else:
write_types += "\n "
write_types += f"b.write(self.{arg_name}.write())\n "
read_types += "\n "
read_types += f"{arg_name} = TLObject.read(b)\n "
slots = ", ".join([f'"{i[0]}"' for i in sorted_args])
return_arguments = ", ".join([f"{i[0]}={i[0]}" for i in sorted_args])
compiled_combinator = combinator_tmpl.format(
notice=notice,
warning=WARNING,
name=c.name,
docstring=docstring,
slots=slots,
id=c.id,
qualname=f"{c.section}.{c.qualname}",
arguments=arguments,
fields=fields,
read_types=read_types,
write_types=write_types,
return_arguments=return_arguments
)
directory = "types" if c.section == "types" else c.section
dir_path = DESTINATION_PATH / directory / c.namespace
os.makedirs(dir_path, exist_ok=True)
module = c.name
if module == "Updates":
module = "UpdatesT"
with open(dir_path / f"{snake(module)}.py", "w") as f:
f.write(compiled_combinator)
d = namespaces_to_constructors if c.section == "types" else namespaces_to_functions
if c.namespace not in d:
d[c.namespace] = []
d[c.namespace].append(c.name)
for namespace, types in namespaces_to_types.items():
with open(DESTINATION_PATH / "base" / namespace / "__init__.py", "w") as f:
f.write(f"{notice}\n\n")
f.write(f"{WARNING}\n\n")
for t in types:
module = t
if module == "Updates":
module = "UpdatesT"
f.write(f"from .{snake(module)} import {t}\n")
if not namespace:
f.write(f"from . import {', '.join(filter(bool, namespaces_to_types))}")
for namespace, types in namespaces_to_constructors.items():
with open(DESTINATION_PATH / "types" / namespace / "__init__.py", "w") as f:
f.write(f"{notice}\n\n")
f.write(f"{WARNING}\n\n")
for t in types:
module = t
if module == "Updates":
module = "UpdatesT"
f.write(f"from .{snake(module)} import {t}\n")
if not namespace:
f.write(f"from . import {', '.join(filter(bool, namespaces_to_constructors))}\n")
for namespace, types in namespaces_to_functions.items():
with open(DESTINATION_PATH / "functions" / namespace / "__init__.py", "w") as f:
f.write(f"{notice}\n\n")
f.write(f"{WARNING}\n\n")
for t in types:
module = t
if module == "Updates":
module = "UpdatesT"
f.write(f"from .{snake(module)} import {t}\n")
if not namespace:
f.write(f"from . import {', '.join(filter(bool, namespaces_to_functions))}")
with open(DESTINATION_PATH / "all.py", "w", encoding="utf-8") as f:
f.write(notice + "\n\n")
f.write(WARNING + "\n\n")
f.write(f"layer = {layer}\n\n")
f.write("objects = {")
for c in combinators:
f.write(f'\n {c.id}: "pyrogram.raw.{c.section}.{c.qualname}",')
f.write('\n 0xbc799737: "pyrogram.raw.core.BoolFalse",')
f.write('\n 0x997275b5: "pyrogram.raw.core.BoolTrue",')
f.write('\n 0x1cb5c415: "pyrogram.raw.core.Vector",')
f.write('\n 0x73f1f8dc: "pyrogram.raw.core.MsgContainer",')
f.write('\n 0xae500895: "pyrogram.raw.core.FutureSalts",')
f.write('\n 0x0949d9dc: "pyrogram.raw.core.FutureSalt",')
f.write('\n 0x3072cfa1: "pyrogram.raw.core.GzipPacked",')
f.write('\n 0x5bb8e511: "pyrogram.raw.core.Message",')
f.write("\n}\n")
if "__main__" == __name__:
HOME_PATH = Path(".")
DESTINATION_PATH = Path("../../pyrogram/raw")
NOTICE_PATH = Path("../../NOTICE")
start(format=False)