v0.9.1

Fix

• Update from older versions (df11e8b)

v0.9.0

Major update

• restructured the code
• cancel, // now remember cancellation permanently
• reactions fix now remembers message content hashes after a restart. This means no more 'Edit message again to evaluate' messages
• implemented simple key-value store for storing various data
  • tgpy.api.config.get(key: str, default: JSON = None) -> JSON
  • tgpy.api.config.set(key: str, value: JSON)
  • tgpy.api.config.unset(key: str)
  • tgpy.api.config.save() useful when modifying objects acquired using the .get method in place
• if the __all__ variable is set in a module, only objects with names in that list are exported (added to variables list)
• ctx.is_module is True if the code is executed as a module (on startup)
• ctx.is_manual_output can be set to True to prevent the last message edit by TGPy so that you can edit it yourself
• cancel, //, restart, ping, update, modules object and .await syntax are now implemented as regular modules in the std directory
  • builtin modules can be disabled with core.disable_modules config option (for example, tgpy.api.config.set('core.disable_modules', ['postfix_await']))
• extra keys are now allowed in module metadata. They are parsed into Module.extra dict
• tgpy.api module is now used for public API instead of the tgpy object
• moved get_installed_version, get_running_version, installed_as_package, get_user, get_hostname, running_in_docker, try_await, outgoing_messages_filter
  functions to tgpy.api (tgpy.api.utils)
• new public API functions:
  • async parse_code(text: str) -> ParseResult(is_code: bool, original: str, transformed: str, tree: AST | None)
  • parse_tgpy_message(message: Message) -> MessageParseResult(is_tgpy_message: bool, code: str | None, result: str | None)
  • async tgpy_eval(code: str, message: Message = None, *, filename: str = None) -> EvalResult(result: Any, output: str)
• AST transformers. AST transformers are applied after code transformers
• exec hooks. Exec hooks are executed before the message is parsed and handled. Exec hooks must have the following signature: async hook(message: Message, is_edit: bool) -> Message | bool | None. An exec hook may edit the message using Telegram API methods and/or alter the message in place. If a hook returns Message object or alters it in place, it's used instead of the original Message object during the rest of handling (including calling other hook functions). If a hook returns True or None, execution completes normally. If a hook returns False, the rest of hooks are executed and then the handling stops without further message parsing or evaluating
  • Apply hooks with tgpy.api.exec_hooks.apply(message: Message, *, is_edit: bool) -> Message | False. This method returns False if any of the hooks returned False, Message object that should be used instead of the original one otherwise
• new dict/list compatible transformers/hooks interface (store_obj is one of tgpy.api.code_transformers, tgpy.api.ast_transformers or tgpy.api.exec_hooks)
  • add a function with store_obj.add(name, function)
  • add/replace a function with store_obj[name] = function
  • get a function with store_obj[name or index]
  • iter functions with for name, function in store_obj
  • apply transformers/hooks with tgpy.api.code_transformers.apply(code: str) -> str, tgpy.api.ast_transformers.apply(tree: AST) -> AST or tgpy.api.exec_hooks.apply (documented above)
• podman is now correctly detected so that tgpy doesn't try to update itself in the container
• Change in-message url to tgpy.tmat.me (8737ca9)
• Move tokenize_string and untokenize_to_string to tgpy.api (7d8c3b2)

v0.8.0

Feature

• Wrap sys.stdout instead of print to capture output + properly use contextvars (7aa2015)

Fix

• Don't stop on unhandled errors. They may happen when, for example, a module uses asyncio.create_task (3584f22)

v0.7.0

Feature

• Update telethon (new markdown parser, html parser fixes) (43dd76f)
• Update dependencies (layer 152) (234dc86)

Fix

• Handle entities properly when editing "//" message (6d989dc)
• Specify parse_mode in error handler to support markdown global parse mode (60cd81b)
• Use message.raw_text instead of message.text to detect // (a85b1c8)

v0.6.2

Fix

• Fix IndentationError appearing for some non-code messages (599c84f)
• docker: Add /venv/bin to path (00be149)

Documentation

• Update readme intro (c334669)

v0.6.1

Fix

• Update command saying "Already up to date" while in fact updating correctly (9b25fe6)

v0.6.0

Feature

• Use MessageEntityPre with language set to 'python' to enable syntax highlighting on supported clients (e.g. WebZ). Closes #24 (5de6ded)

Fix

• Keep 'cancel' message when replying to other user (fixes #21) (057231d)
• Ignore error when running code deletes the message itself (d022450)

Documentation

• Update readme (a321bbf)

v0.5.1

Fix

• Restart() now edits message properly (60afa44)
• Compatibility with python 3.9 (f3c0468)

v0.5.0

Feature

• Use custom telethon fork with updated layer (91894fc) (9c7738e)
• Transform x.await into await x (6117421) (fde6291)
• Better version detection, ping() builtin improvements (265b83f)
• Allow to specify data directory via environment variable TGPY_DATA (4d769da)

Fix

• Setting/removing reaction no longer triggers reevaluation (#25) (cf6e64e)
• Emojis no longer break messages markup (#10) (20f48bc)
• Parsing of modules with triple quotes in code (485166d)

Documentation

• Guide rewrite

v0.4.1

Fix

• code detection: Ignore messages like "fix: fix" (1b73815)