Update code generator to parse CSV files

setup.py

@@ -15,7 +15,6 @@ import json
 import os
 import re
 import shutil
-from codecs import open
 from sys import argv
 
 from setuptools import find_packages, setup
@@ -40,11 +39,10 @@ class TempWorkDir:
 GENERATOR_DIR = 'telethon_generator'
 LIBRARY_DIR = 'telethon'
 
-ERRORS_IN_JSON = os.path.join(GENERATOR_DIR, 'data', 'errors.json')
-ERRORS_IN_DESC = os.path.join(GENERATOR_DIR, 'data', 'error_descriptions')
+ERRORS_IN = os.path.join(GENERATOR_DIR, 'data', 'errors.csv')
 ERRORS_OUT = os.path.join(LIBRARY_DIR, 'errors', 'rpcerrorlist.py')
 
-INVALID_BM_IN = os.path.join(GENERATOR_DIR, 'data', 'invalid_bot_methods.json')
+METHODS_IN = os.path.join(GENERATOR_DIR, 'data', 'methods.csv')
 
 TLOBJECT_IN_CORE_TL = os.path.join(GENERATOR_DIR, 'data', 'mtproto_api.tl')
 TLOBJECT_IN_TL = os.path.join(GENERATOR_DIR, 'data', 'telegram_api.tl')
@@ -56,16 +54,19 @@ DOCS_OUT = 'docs'
 
 
 def generate(which):
-    from telethon_generator.parsers import parse_errors, parse_tl, find_layer
+    # TODO make docs generator use the new CSV too
+    from telethon_generator.parsers import\
+        parse_errors, parse_methods, parse_tl, find_layer
+
     from telethon_generator.generators import\
         generate_errors, generate_tlobjects, generate_docs, clean_tlobjects
 
-    # Older Python versions open the file as bytes instead (3.4.2)
-    with open(INVALID_BM_IN, 'r') as f:
-        invalid_bot_methods = set(json.load(f))
-
     layer = find_layer(TLOBJECT_IN_TL)
-    errors = list(parse_errors(ERRORS_IN_JSON, ERRORS_IN_DESC))
+    errors = list(parse_errors(ERRORS_IN))
+    methods = list(parse_methods(METHODS_IN, {e.str_code: e for e in errors}))
+    invalid_bot_methods = {m.name for m in methods
+                           if not m.usability.startswith('bot')}
+
     tlobjects = list(itertools.chain(
         parse_tl(TLOBJECT_IN_CORE_TL, layer, invalid_bot_methods),
         parse_tl(TLOBJECT_IN_TL, layer, invalid_bot_methods)))

telethon_generator/generators/docs.py

@@ -1,4 +1,5 @@
 #!/usr/bin/env python3
+import csv
 import functools
 import os
 import re
@@ -9,7 +10,6 @@ from ..docswriter import DocsWriter
 from ..parsers import TLObject
 from ..utils import snake_to_camel_case
 
-
 CORE_TYPES = {
     'int', 'long', 'int128', 'int256', 'double',
     'vector', 'string', 'bool', 'true', 'bytes', 'date'

telethon_generator/parsers/__init__.py

@@ -1,2 +1,3 @@
 from .errors import Error, parse_errors
+from .methods import MethodInfo, parse_methods
 from .tlobject import TLObject, parse_tl, find_layer

telethon_generator/parsers/errors.py

@@ -1,6 +1,4 @@
-import json
-import re
-from collections import defaultdict
+import csv
 
 from ..utils import snake_to_camel_case
 
@@ -16,15 +14,8 @@ KNOWN_BASE_CLASSES = {
     500: 'ServerError',
 }
 
-# The API doesn't return the code for some (vital) errors. They are
-# all assumed to be 400, except these well-known ones that aren't.
-KNOWN_CODES = {
-    'ACTIVE_USER_REQUIRED': 401,
-    'AUTH_KEY_UNREGISTERED': 401,
-    'USER_DEACTIVATED': 401
-}
-
 # Give better semantic names to some captures
+# TODO Move this to the CSV?
 CAPTURE_NAMES = {
     'FloodWaitError': 'seconds',
     'FloodTestPhoneWaitError': 'seconds',
@@ -52,96 +43,40 @@ def _get_class_name(error_code):
 
 
 class Error:
-    def __init__(self, int_code, str_code, description, caused_by):
-        # TODO Some errors have the same str_code but different int_code
+    def __init__(self, codes, name, description):
+        # TODO Some errors have the same name but different integer codes
         # Should these be split into different files or doesn't really matter?
         # Telegram isn't exactly consistent with returned errors anyway.
-        self.int_code = int_code
-        self.str_code = str_code
-        self.subclass = _get_class_name(int_code)
-        self.subclass_exists = int_code in KNOWN_BASE_CLASSES
+        self.int_code = codes[0]
+        self.str_code = name
+        self.subclass = _get_class_name(codes[0])
+        self.subclass_exists = codes[0] in KNOWN_BASE_CLASSES
         self.description = description
-        self.caused_by = list(sorted(caused_by))
 
-        self.has_captures = '_X' in str_code
+        self.has_captures = '_X' in name
         if self.has_captures:
-            self.name = _get_class_name(str_code.replace('_X', ''))
-            self.pattern = str_code.replace('_X', r'_(\d+)')
+            self.name = _get_class_name(name.replace('_X', ''))
+            self.pattern = name.replace('_X', r'_(\d+)')
             self.capture_name = CAPTURE_NAMES.get(self.name, 'x')
         else:
-            self.name = _get_class_name(str_code)
-            self.pattern = str_code
+            self.name = _get_class_name(name)
+            self.pattern = name
             self.capture_name = None
 
 
-def parse_errors(json_file, descriptions_file):
+def parse_errors(csv_file):
     """
-    Parses the given JSON file in the following format:
-        {
-            "ok": true,
-            "human_result": {"int_code": ["descriptions"]},
-            "result": {"int_code": {"full_method_name": ["str_error"]}}
-        }
-
-    The descriptions file, which has precedence over the JSON's human_result,
-    should have the following format:
-        # comment
-        str_error=Description
-
-    The method yields `Error` instances as a result.
+    Parses the input CSV file with columns (name, error codes, description)
+    and yields `Error` instances as a result.
     """
-    with open(json_file, 'r', encoding='utf-8') as f:
-        data = json.load(f)
+    with open(csv_file, newline='') as f:
+        f = csv.reader(f)
+        next(f, None)  # header
+        for line, (name, codes, description) in enumerate(f, start=2):
+            try:
+                codes = [int(x) for x in codes.split()] or [400]
+            except ValueError:
+                raise ValueError('Not all codes are integers '
+                                 '(line {})'.format(line)) from None
 
-    errors = defaultdict(set)
-    error_to_method = defaultdict(set)
-    # PWRTelegram's API doesn't return all errors, which we do need here.
-    # Add some special known-cases manually first.
-    errors[420].update((
-        'FLOOD_WAIT_X', 'FLOOD_TEST_PHONE_WAIT_X', 'TAKEOUT_INIT_DELAY_X'
-    ))
-    errors[401].update((
-        'AUTH_KEY_INVALID', 'SESSION_EXPIRED', 'SESSION_REVOKED'
-    ))
-    errors[303].update((
-        'FILE_MIGRATE_X', 'PHONE_MIGRATE_X',
-        'NETWORK_MIGRATE_X', 'USER_MIGRATE_X'
-    ))
-    for int_code, method_errors in data['result'].items():
-        for method, error_list in method_errors.items():
-            for error in error_list:
-                error = re.sub('_\d+', '_X', error).upper()
-                errors[int(int_code)].add(error)
-                error_to_method[error].add(method)
-
-    # Some errors are in the human result, but not with a code. Assume 400
-    for error in data['human_result']:
-        if error[0] != '-' and not error.isdigit():
-            error = re.sub('_\d+', '_X', error).upper()
-            if not any(error in es for es in errors.values()):
-                errors[KNOWN_CODES.get(error, 400)].add(error)
-
-    # Prefer the descriptions that are related with Telethon way of coding
-    # to those that PWRTelegram's API provides.
-    telethon_descriptions = {}
-    with open(descriptions_file, 'r', encoding='utf-8') as f:
-        for line in f:
-            line = line.strip()
-            if line and not line.startswith('#'):
-                equal = line.index('=')
-                message, description = line[:equal], line[equal + 1:]
-                telethon_descriptions[message.rstrip()] = description.lstrip()
-
-    for int_code, error_set in errors.items():
-        for str_code in sorted(error_set):
-            description = telethon_descriptions.get(
-                str_code, '\n'.join(data['human_result'].get(
-                    str_code, ['No description known']
-                ))
-            )
-            yield Error(
-                int_code=int_code,
-                str_code=str_code,
-                description=description,
-                caused_by=error_to_method[str_code]
-            )
+            yield Error([int(x) for x in codes], name, description)

telethon_generator/parsers/methods.py

@@ -0,0 +1,29 @@
+import csv
+
+
+class MethodInfo:
+    def __init__(self, name, usability, errors):
+        self.name = name
+        self.usability = usability
+        self.errors = errors
+
+
+def parse_methods(csv_file, errors_dict):
+    """
+    Parses the input CSV file with columns (method, usability, errors)
+    and yields `MethodInfo` instances as a result.
+    """
+    with open(csv_file, newline='') as f:
+        f = csv.reader(f)
+        next(f, None)  # header
+        for line, (method, usability, errors) in enumerate(f, start=2):
+            if usability not in ('user', 'bot', 'both', 'unknown'):
+                raise ValueError('Usability must be either user, bot, '
+                                 'both or unknown, not {}'.format(usability))
+            try:
+                errors = [errors_dict[x] for x in errors.split()]
+            except KeyError:
+                raise ValueError('Method {} references unknown errors {}'
+                                 .format(method, errors)) from None
+
+            yield MethodInfo(method, usability, errors)