PyPtt 2.2.4__tar.gz → 2.2.6__tar.gz

{pyptt-2.2.4 → pyptt-2.2.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: PyPtt
-Version: 2.2.4
+Version: 2.2.6
 Summary: PyPtt github: https://github.com/PyPtt/PyPtt
 Author-email: CodingMan <pttcodingman@gmail.com>
 License: GNU Lesser General Public License v3 (LGPLv3)

{pyptt-2.2.4 → pyptt-2.2.6}/PyPtt/PTT.py

@@ -28,6 +28,7 @@ from . import _api_post
 from . import _api_reply_post
 from . import _api_search_user
 from . import _api_set_board_title
+from . import _api_set_signature_file
 from . import check_value
 from . import config
 from . import connect_core
@@ -806,6 +807,38 @@ class API:
 
         _api_set_board_title.set_board_title(self, board, new_title)
 
+    def set_signature_file(self, content: str) -> None:
+
+        """
+        更新登入帳號的名片檔（PTT 的 plan），對應【個人設定】->【個人檔案】
+        ->(Q)ueryEdit 編輯名片檔的功能。`get_user()` 回傳的 `signature_file`
+        欄位讀的就是這份內容。
+
+        Args:
+            content (str): 新的名片檔內容，會整份取代原本的內容。
+
+        Returns:
+            None
+
+        Raises:
+            RequireLogin: 需要登入。
+
+        範例::
+
+            import PyPtt
+
+            ptt_bot = PyPtt.API()
+            try:
+                # .. login ..
+                ptt_bot.set_signature_file(content='你好，我是 PyPtt')
+                # .. do something ..
+            finally:
+                ptt_bot.logout()
+
+        """
+
+        _api_set_signature_file.set_signature_file(self, content)
+
     def mark_post(self, mark_type: int, board: str, aid: Optional[str] = None, index: int = 0, search_type: int = 0,
                   search_condition: Optional[str] = None) -> None:
         """

{pyptt-2.2.4 → pyptt-2.2.6}/PyPtt/__init__.py

@@ -1,4 +1,4 @@
-__version__ = '2.2.4'
+__version__ = '2.2.6'
 __author__ = 'CodingMan'
 __email__ = 'pttcodingman@gmail.com'
 

pyptt-2.2.6/PyPtt/_api_set_signature_file.py

@@ -0,0 +1,86 @@
+from . import _api_util
+from . import check_value
+from . import command
+from . import connect_core
+from . import exceptions
+from . import i18n
+from . import lib_util
+from . import log
+
+
+def set_signature_file(api, content: str) -> None:
+    _api_util.one_thread(api)
+
+    if not api._is_login:
+        raise exceptions.RequireLogin(i18n.require_login)
+
+    check_value.check_type(content, str, 'content')
+
+    log.logger.info(i18n.set_signature_file)
+
+    cmd_list = []
+    cmd_list.append(command.go_main_menu)
+    cmd_list.append('U')
+    cmd_list.append(command.enter)
+    cmd_list.append('M')
+    cmd_list.append(command.enter)
+    cmd_list.append('Q')
+    cmd_list.append(command.enter)
+    cmd = ''.join(cmd_list)
+
+    target_list = [
+        # Existing signature file -> getans asks Delete/Edit/Cancel first.
+        # max_match=1: the echoed prompt text can linger on the next
+        # screen refresh while waiting for the editor to draw (same
+        # quirk documented for '確定要儲存檔案嗎' in _api_post.py); without
+        # it this would re-fire 'E' + Enter into the editor buffer.
+        connect_core.TargetUnit('名片 (D)刪除', response='E' + command.enter, max_match=1),
+        # Empty signature file skips the getans prompt and opens the
+        # editor directly.
+        connect_core.TargetUnit('編輯文章', break_detect=True),
+    ]
+    index = api.connect_core.send(
+        cmd,
+        target_list,
+        screen_timeout=api.config.screen_long_timeout)
+    if index < 0:
+        ori_screen = api.connect_core.get_screen_queue()[-1]
+        raise exceptions.UnknownError(ori_screen)
+
+    content = lib_util.uniform_new_line(content)
+
+    cmd_list = []
+    # ponytail: unlike post's blank template, the card-file editor preloads
+    # the account's existing signature file, so it has to be cleared line by
+    # line before writing the new content. ctrl_y*60 is a "clear up to 60
+    # pre-loaded lines" ceiling -- bump it if a real card file ever runs
+    # longer than that.
+    cmd_list.append(command.ctrl_y * 60)
+    cmd_list.append(content)
+    cmd_list.append(command.ctrl_x)
+    cmd = ''.join(cmd_list)
+
+    target_list = [
+        connect_core.TargetUnit('確定要儲存檔案嗎', response='s' + command.enter, max_match=1),
+        connect_core.TargetUnit('名片更新完畢', break_detect=True),
+    ]
+    index = api.connect_core.send(
+        cmd,
+        target_list,
+        screen_timeout=api.config.screen_long_timeout)
+    if index < 0:
+        ori_screen = api.connect_core.get_screen_queue()[-1]
+        raise exceptions.UnknownError(ori_screen)
+
+    # The leading space in go_main_menu dismisses the "請按任意鍵繼續" banner
+    # left by the save; the trailing left-arrows walk back out of the
+    # 個人設定/個人檔案 submenus to the main menu.
+    target_list = [
+        connect_core.TargetUnit('主功能表', break_detect=True),
+    ]
+    api.connect_core.send(
+        command.go_main_menu,
+        target_list,
+        screen_timeout=api.config.screen_long_timeout)
+
+    log.logger.info(i18n.set_signature_file, '...', i18n.success)

{pyptt-2.2.4 → pyptt-2.2.6}/PyPtt/connect_core.py

@@ -160,6 +160,7 @@ class API(object):
         self._UseTooManyResources = TargetUnit(screens.Target.use_too_many_resources,
                                                exceptions_=exceptions.UseTooManyResources())
         self._loop = None
+        self._stream_parsers = {}
         self._ssl_context = ssl_init(self.config.verify_ssl)
 
     def _get_event_loop(self):
@@ -257,14 +258,11 @@ class API(object):
         if not connect_success:
             raise exceptions.ConnectError(self.config)
 
-    def _decode_screen(self, receive_data_buffer, start_time, target_list, is_secret, refresh, msg):
+    def _decode_screen(self, screen, start_time, target_list, is_secret, refresh, msg):
 
         break_detect_after_send = False
         use_too_many_res = False
 
-        vt100_p = screens.VT100Parser(receive_data_buffer, self.current_encoding, self.config.screen_height)
-        screen = vt100_p.screen
-
         find_target = False
         target_index = -1
         for target in target_list:
@@ -308,6 +306,17 @@ class API(object):
                 break
         return screen, find_target, is_secret, break_detect_after_send, use_too_many_res, msg, target_index
 
+    def _stream_screen(self, encoding: str, data_chunk: bytes) -> str:
+        """Feed a newly-received chunk to the per-encoding incremental parser
+        and return the current screen. Each parser decodes and steps through
+        every byte exactly once across the whole receive sequence."""
+        parser = self._stream_parsers.get(encoding)
+        if parser is None:
+            parser = screens.IncrementalScreen(encoding, self.config.screen_height)
+            self._stream_parsers[encoding] = parser
+        parser.feed(data_chunk)
+        return parser.screen
+
     async def _async_send(self, msg: str, target_list: list, screen_timeout: int, refresh: bool, secret: bool) -> int:
         current_screen_timeout = self.config.screen_timeout if screen_timeout == 0 else screen_timeout
         is_secret = secret
@@ -342,6 +351,11 @@ class API(object):
 
             msg = ''
             receive_data_buffer = bytes()
+            # Fresh incremental parsers for this screen sequence. Each arriving
+            # chunk is fed once (per encoding) instead of re-parsing the whole
+            # accumulated buffer every time — turning the receive loop from
+            # O(N²) into O(N).
+            self._stream_parsers = {}
             start_time = time.time()
             find_target = False
             target_index = -1
@@ -360,14 +374,20 @@ class API(object):
                             data_chunk = data_chunk.encode('utf-8')
                         receive_data_buffer += data_chunk
 
+                        screen = self._stream_screen(self.current_encoding, data_chunk)
                         screen, find_target, is_secret, break_detect_after_send, use_too_many_res, msg, target_index = \
-                            self._decode_screen(receive_data_buffer, start_time, target_list, is_secret, refresh, msg)
+                            self._decode_screen(screen, start_time, target_list, is_secret, refresh, msg)
 
                         if not find_target:
+                            # Auto-detect the server encoding: the alternate
+                            # parser has also seen every chunk so far (this
+                            # branch runs on every non-matching chunk), so
+                            # feeding it this chunk keeps it in sync.
                             original_encoding = self.current_encoding
                             self.current_encoding = 'big5uao' if original_encoding == 'utf-8' else 'utf-8'
+                            screen_ = self._stream_screen(self.current_encoding, data_chunk)
                             screen_, find_target, is_secret, break_detect_after_send, use_too_many_res, msg, target_index = \
-                                self._decode_screen(receive_data_buffer, start_time, target_list, is_secret, refresh, msg)
+                                self._decode_screen(screen_, start_time, target_list, is_secret, refresh, msg)
                             if find_target:
                                 screen = screen_
                             else:
@@ -431,6 +451,7 @@ class API(object):
                     return -1
                 msg = ''
                 receive_data_buffer = bytes()
+                self._stream_parsers = {}
                 start_time = time.time()
                 mid_time = time.time()
                 while mid_time - start_time < current_screen_timeout:
@@ -439,8 +460,9 @@ class API(object):
                     except EOFError:
                         return -1
                     receive_data_buffer += data
+                    screen = self._stream_screen(self.current_encoding, data)
                     screen, find_target, is_secret, break_detect_after_send, use_too_many_res, msg, target_index = \
-                        self._decode_screen(receive_data_buffer, start_time, target_list, is_secret, refresh, msg)
+                        self._decode_screen(screen, start_time, target_list, is_secret, refresh, msg)
                     if target_index != -1:
                         return target_index
                     if use_too_many_res:

{pyptt-2.2.4 → pyptt-2.2.6}/PyPtt/lang_en_US.py

@@ -106,6 +106,7 @@ string_data = {
     "set_connect_host": "Set up the connect host",
     "set_connect_mode": "Set up the connect mode",
     "set_contact_mail_first": "Password can only be changed after setting the contact mailbox",
+    "set_signature_file": "Update signature file",
     "set_up_lang_module": "Set up language module",
     "spend_time": "Spend time",
     "substandard_post": "Substandard post",

{pyptt-2.2.4 → pyptt-2.2.6}/PyPtt/lang_zh_TW.py

@@ -106,6 +106,7 @@ string_data = {
     "set_connect_host": "設定連線主機",
     "set_connect_mode": "設定連線模式",
     "set_contact_mail_first": "請先設定聯絡信箱後才能修改密碼",
+    "set_signature_file": "更新名片檔",
     "set_up_lang_module": "設定語言模組",
     "spend_time": "花費時間",
     "substandard_post": "不合規範文章",

{pyptt-2.2.4 → pyptt-2.2.6}/PyPtt/screens.py

@@ -1,3 +1,4 @@
+import codecs
 import re
 import sys
 import unicodedata
@@ -9,6 +10,14 @@ from . import log
 register_uao()
 
 
+# Memoise the per-character width. east_asian_width is a C call, but a PTT
+# screen is a few thousand characters drawn from a small alphabet, and the
+# parser walks each line repeatedly (_str_pos_at_cells, replace mode). Caching
+# turns the hot path into a dict hit. Writes are idempotent, so the plain-dict
+# race under free-threading is benign.
+_cell_width_cache: dict = {}
+
+
 def _cell_width(ch: str) -> int:
     """Return the terminal-cell width of a single character.
 
@@ -17,7 +26,11 @@ def _cell_width(ch: str) -> int:
     Ambiguous includes glyphs like ※, ←, →, ◎ that PTT's protocol treats
     as full-width when positioning the cursor.
     """
-    return 2 if unicodedata.east_asian_width(ch) in ('W', 'F', 'A') else 1
+    w = _cell_width_cache.get(ch)
+    if w is None:
+        w = 2 if unicodedata.east_asian_width(ch) in ('W', 'F', 'A') else 1
+        _cell_width_cache[ch] = w
+    return w
 
 
 def _cell_len(text: str) -> int:
@@ -241,51 +254,76 @@ def show(config, screen_queue, function_name=None):
 
 xy_pattern_h = re.compile(r'^=ESC=\[[\d]+;[\d]+H')
 xy_pattern_s = re.compile(r'^=ESC=\[[\d]+;[\d]+s')
+_color_sgr = re.compile(r'\x1B\[[\d+;]*m')
+
+
+def _preprocess(data: str) -> str:
+    """Normalise a decoded chunk: strip color, mark escapes, drop CR, and
+    collapse PTT's ' \\x08' wide-char reservation pairs.
+
+    Only the SGR (color) strip genuinely needs a regex; the remaining
+    substitutions are literal, and str.replace is several times faster than
+    re.sub — this runs on the whole buffer on every parse."""
+    data = _color_sgr.sub('', data)
+    data = data.replace('\x1B', '=ESC=')
+    data = data.replace('\r', '')
+    while ' \x08' in data:
+        data = data.replace(' \x08', '')
+    return data
+
+
+class _VT100Engine:
+    """Shared VT100 stepping engine.
+
+    Both the batch ``VT100Parser`` and the streaming ``IncrementalScreen``
+    drive the same left-to-right state machine over a preprocessed string
+    (escapes marked as ``=ESC=``). The screen is tracked in terminal cells
+    (see the module-level cell-width helpers), not bytes.
 
+    ``ESC [ 2J`` (clear screen) is handled as an in-loop reset: processing
+    left-to-right with a reset at every 2J yields the same final screen as
+    the classic "discard everything before the last 2J" slice, because each
+    reset wipes prior state — only work after the final reset survives."""
 
-class VT100Parser:
+    def __init__(self, screen_height: int = 24):
+        self._height = screen_height
+        self._reset_screen()
+
+    def _reset_screen(self):
+        self._cursor_x = 0
+        self._cursor_y = 0
+        self._lines = [''] * self._height
+        self._screen_length = dict()
+
+    # ── cursor / erase primitives ─────────────────────────────────────────
     def _h(self):
         self._cursor_x = 0
         self._cursor_y = 0
 
     def _move(self, x, y):
         self._cursor_x = x
-        self._cursor_y = min(y, len(self.screen) - 1)
+        self._cursor_y = min(y, len(self._lines) - 1)
 
     def _newline(self):
         self._cursor_x = 0
-        self._cursor_y = min(self._cursor_y + 1, len(self.screen) - 1)
+        self._cursor_y = min(self._cursor_y + 1, len(self._lines) - 1)
 
     def _k(self):
         # Erase from cursor to end of line (VT100 ESC [ K).
-        pos = _str_pos_at_cells(self.screen[self._cursor_y], self._cursor_x)
-        self.screen[self._cursor_y] = self.screen[self._cursor_y][:pos]
-        self.screen_length[self._cursor_y] = self._cursor_x
-
-    def __init__(self, bytes_data, encoding, screen_height: int = 24):
-        # self._data = data
-        # https://www.csie.ntu.edu.tw/~r88009/Java/html/Network/vt100.htm
-
-        self._cursor_x = 0
-        self._cursor_y = 0
-        self.screen = [''] * screen_height
-        self.screen_length = dict()
-
-        data = bytes_data.decode(encoding, errors='replace')
-
-        # remove color
-        data = re.sub(r'\x1B\[[\d+;]*m', '', data)
-        data = re.sub(r'[\x1B]', '=ESC=', data)
-        data = re.sub(r'[\r]', '', data)
-        while ' \x08' in data:
-            data = re.sub(r' \x08', '', data)
-
-        if '=ESC=[2J' in data:
-            data = data[data.rfind('=ESC=[2J') + len('=ESC=[2J'):]
-
-        count = 0
+        pos = _str_pos_at_cells(self._lines[self._cursor_y], self._cursor_x)
+        self._lines[self._cursor_y] = self._lines[self._cursor_y][:pos]
+        self._screen_length[self._cursor_y] = self._cursor_x
+
+    # ── the stepping loop ─────────────────────────────────────────────────
+    def _run(self, data: str) -> str:
+        """Consume `data` left-to-right, mutating screen state.
+
+        Returns the unconsumed tail: '' when everything was consumed, or the
+        remainder starting at a front ``=ESC=`` the machine does not
+        recognise. The batch parser ignores that tail (matching the classic
+        "bail on unknown escape" behaviour); the incremental parser carries
+        it until more bytes arrive or freezes on it."""
         while data:
-            count += 1
             while True:
                 if not data.startswith('=ESC='):
                     break
@@ -300,6 +338,10 @@ class VT100Parser:
                 elif data.startswith('=ESC=[s'):
                     data = data[len('=ESC=[s'):]
                     continue
+                elif data.startswith('=ESC=[2J'):
+                    data = data[len('=ESC=[2J'):]
+                    self._reset_screen()
+                    continue
                 break
 
             xy_result = None
@@ -317,7 +359,6 @@ class VT100Parser:
                 new_y = int(xy_part[6:xy_part.find(';')]) - 1
                 # VT100 columns are 1-based; convert to 0-based cursor_x.
                 new_x = int(xy_part[xy_part.find(';') + 1: -1]) - 1
-                # log.py.info('xy', xy_part, new_x, new_y)
                 self._move(new_x, new_y)
 
                 data = data[len(xy_part):]
@@ -328,16 +369,14 @@ class VT100Parser:
                     self._newline()
                     continue
 
-                # print(f'-{data[:1]}-{len(data[:1].encode("big5-uao", "replace"))}')
+                if self._cursor_y not in self._screen_length:
+                    self._screen_length[self._cursor_y] = _cell_len(self._lines[self._cursor_y])
 
-                if self._cursor_y not in self.screen_length:
-                    self.screen_length[self._cursor_y] = _cell_len(self.screen[self._cursor_y])
-
-                current_line_length = self.screen_length[self._cursor_y]
+                current_line_length = self._screen_length[self._cursor_y]
                 replace_mode = False
                 if current_line_length < self._cursor_x:
                     append_space = ' ' * (self._cursor_x - current_line_length)
-                    self.screen[self._cursor_y] += append_space
+                    self._lines[self._cursor_y] += append_space
                 elif current_line_length > self._cursor_x:
                     replace_mode = True
 
@@ -347,34 +386,143 @@ class VT100Parser:
                 next_esc = data.find('=ESC=')
                 next_esc = 1920 if next_esc < 0 else next_esc
                 if next_esc == 0:
-                    break
+                    # A front escape the machine does not consume. Hand the
+                    # remainder back to the caller instead of dropping it.
+                    return data
 
                 current_index = min(next_newline, next_esc)
 
                 current_data = data[:current_index]
                 current_data_length = _cell_len(current_data)
                 if replace_mode:
-                    line = self.screen[self._cursor_y]
+                    line = self._lines[self._cursor_y]
                     splice_start = _str_pos_at_cells(line, self._cursor_x)
                     splice_end = _str_pos_at_cells(line, self._cursor_x + current_data_length)
-                    self.screen[self._cursor_y] = line[:splice_start] + current_data + line[splice_end:]
+                    self._lines[self._cursor_y] = line[:splice_start] + current_data + line[splice_end:]
                     self._cursor_x += current_data_length
-                    if self._cursor_x > self.screen_length[self._cursor_y]:
-                        self.screen_length[self._cursor_y] = self._cursor_x
+                    if self._cursor_x > self._screen_length[self._cursor_y]:
+                        self._screen_length[self._cursor_y] = self._cursor_x
                 else:
-                    self.screen[self._cursor_y] += current_data
+                    self._lines[self._cursor_y] += current_data
                     self._cursor_x += current_data_length
-                    self.screen_length[self._cursor_y] = self._cursor_x
+                    self._screen_length[self._cursor_y] = self._cursor_x
 
                 data = data[current_index:]
+        return ''
+
+
+class VT100Parser(_VT100Engine):
+    """Batch parser: decode the whole byte buffer and render one screen.
 
-                # print('\n'.join(self.screen))
-        # print('\n'.join(self._screen))
-        # print('=' * 20)
-        # print(data)
+    Kept as the reference implementation and the target-matching path for a
+    full (non-streamed) buffer. ``IncrementalScreen`` fed the same bytes in
+    any chunking produces a byte-identical screen (see
+    tests/test_incremental_parity.py)."""
 
-        # print('Spend', count, 'cycle')
-        self.screen = '\n'.join(self.screen)
+    def __init__(self, bytes_data, encoding, screen_height: int = 24):
+        # https://www.csie.ntu.edu.tw/~r88009/Java/html/Network/vt100.htm
+        super().__init__(screen_height)
+        data = _preprocess(bytes_data.decode(encoding, errors='replace'))
+        # Leftover (a trailing/unknown escape) is intentionally ignored here,
+        # matching the classic parser's "bail on unknown escape" behaviour.
+        self._run(data)
+        self.screen = '\n'.join(self._lines)
+
+
+# CSI final bytes are 0x40-0x7e; parameter/intermediate bytes are 0x20-0x3f.
+def _escape_complete(seq: str) -> bool:
+    """Is `seq` (which starts with a raw ESC) a complete escape sequence?
+
+    Used by the streaming parser to decide whether a trailing ESC run might
+    still be extended by the next chunk. A CSI (``ESC [``) is complete once a
+    final byte (0x40-0x7e) appears; a non-CSI ESC is treated as complete (the
+    engine will bail on it, same as the batch parser)."""
+    if len(seq) < 2:
+        return False            # lone ESC — wait for more
+    if seq[1] != '[':
+        return True             # ESC + non-'[': engine bails on it
+    for ch in seq[2:]:
+        if 0x40 <= ord(ch) <= 0x7e:
+            return True
+    return False                # still inside CSI parameters
+
+
+class IncrementalScreen(_VT100Engine):
+    """Streaming VT100 parser: feed byte chunks, read the current screen.
+
+    Removes the O(N²) re-parse of the receive loop — each byte is decoded and
+    stepped through the state machine exactly once, instead of the whole
+    accumulated buffer being re-parsed on every arriving chunk.
+
+    Correctness is anchored to ``VT100Parser``: feeding the same bytes in any
+    chunking (down to one byte at a time) yields the identical screen."""
+
+    def __init__(self, encoding, screen_height: int = 24):
+        super().__init__(screen_height)
+        self._decoder = codecs.getincrementaldecoder(encoding)(errors='replace')
+        self._raw = ''          # decoded but not-yet-safe-to-process tail
+        self._leftover = ''     # preprocessed front escape the engine paused on
+        self._bailed = False    # hit a complete-but-unknown escape → frozen
+
+    def _safe_cut(self, s: str) -> int:
+        """Length of the prefix of `s` safe to process now.
+
+        Holds back the maximal trailing run of spaces and backspaces, and an
+        incomplete trailing escape.
+
+        The ' \\x08' collapse cascades backwards through a run of spaces
+        (``'   \\x08\\x08'`` → ``' '``), and the ``\\x08`` bytes can arrive in a
+        later chunk, so the whole trailing ``[ \\x08]`` run is ambiguous until a
+        character that cannot participate in the collapse follows it. Only
+        commit up to that terminating character."""
+        cut = len(s)
+        while cut > 0 and s[cut - 1] in ' \x08':
+            cut -= 1
+        esc = s.rfind('\x1b')
+        if esc != -1 and not _escape_complete(s[esc:]):
+            cut = min(cut, esc)
+        return cut
+
+    def feed(self, chunk) -> None:
+        if self._bailed:
+            return
+        if isinstance(chunk, str):
+            chunk = chunk.encode('utf-8')
+        self._raw += self._decoder.decode(chunk)
+        cut = self._safe_cut(self._raw)
+        if cut == 0:
+            return
+        safe = _preprocess(self._raw[:cut])
+        self._raw = self._raw[cut:]
+        leftover = self._run(self._leftover + safe)
+        self._leftover = leftover
+        if leftover:
+            # _safe_cut only forwards complete escapes, so a surviving front
+            # escape is complete but unrecognised — the batch parser bails on
+            # it forever, so we freeze to match.
+            self._bailed = True
+
+    @property
+    def screen(self) -> str:
+        """Current screen as a ``screen_height``-line string, including any
+        still-buffered tail — matches VT100Parser over all bytes fed so far."""
+        dec_state = self._decoder.getstate()
+        if not self._raw and not self._leftover and not dec_state[0]:
+            # Nothing held and no partial multibyte in the decoder.
+            return '\n'.join(self._lines)
+        # Tentatively flush the held tail on a snapshot so a partial escape, an
+        # unpaired trailing space, or a partial multibyte char is reflected
+        # (matching the batch parser's errors='replace') without committing.
+        saved = (self._cursor_x, self._cursor_y, list(self._lines),
+                 dict(self._screen_length))
+        try:
+            tail = self._raw + self._decoder.decode(b'', final=True)
+            self._run(self._leftover + _preprocess(tail))
+            return '\n'.join(self._lines)
+        finally:
+            (self._cursor_x, self._cursor_y, self._lines,
+             self._screen_length) = saved
+            self._decoder.setstate(dec_state)
 
 
 if __name__ == '__main__':

{pyptt-2.2.4 → pyptt-2.2.6}/PyPtt.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: PyPtt
-Version: 2.2.4
+Version: 2.2.6
 Summary: PyPtt github: https://github.com/PyPtt/PyPtt
 Author-email: CodingMan <pttcodingman@gmail.com>
 License: GNU Lesser General Public License v3 (LGPLv3)

{pyptt-2.2.4 → pyptt-2.2.6}/PyPtt.egg-info/SOURCES.txt

@@ -30,6 +30,7 @@ PyPtt/_api_post.py
 PyPtt/_api_reply_post.py
 PyPtt/_api_search_user.py
 PyPtt/_api_set_board_title.py
+PyPtt/_api_set_signature_file.py
 PyPtt/_api_util.py
 PyPtt/api_server.py
 PyPtt/check_value.py
@@ -78,6 +79,7 @@ tests/test_get_user.py
 tests/test_get_waterball.py
 tests/test_give_money.py
 tests/test_i18n.py
+tests/test_incremental_parity.py
 tests/test_init.py
 tests/test_logger.py
 tests/test_login_logout.py
@@ -92,4 +94,5 @@ tests/test_reply_post.py
 tests/test_search_user.py
 tests/test_service.py
 tests/test_set_board_title.py
+tests/test_set_signature_file.py
 tests/test_vt100_parser.py

pyptt-2.2.6/tests/test_incremental_parity.py

@@ -0,0 +1,135 @@
+"""
+Parity tests for PyPtt.screens.IncrementalScreen.
+
+The receive loop in connect_core streams bytes and must decide, after every
+arriving chunk, whether the current screen matches a target. Re-parsing the
+whole accumulated buffer on every chunk is O(N²); IncrementalScreen feeds each
+byte through the VT100 state machine exactly once instead.
+
+Correctness is anchored to the batch VT100Parser: feeding the same bytes in any
+chunking — down to one byte at a time, the worst case for split escapes, split
+multibyte characters and split ' \\x08' wide-char pairs — must yield the exact
+same screen string. These tests pin that invariant so a future change to either
+parser can't silently diverge.
+"""
+
+import glob
+import random
+from pathlib import Path
+
+import pytest
+
+from PyPtt.screens import IncrementalScreen, VT100Parser
+
+FIXTURE_DIR = Path(__file__).resolve().parent / 'fixtures' / 'vt100'
+
+CHUNK_SIZES = [1, 2, 3, 5, 7, 13, 64, 256, 100000]
+
+
+def feed_in_chunks(data: bytes, encoding: str, chunk: int,
+                   read_each: bool = False) -> str:
+    parser = IncrementalScreen(encoding, 24)
+    for i in range(0, len(data), chunk):
+        parser.feed(data[i:i + chunk])
+        if read_each:
+            # Reading .screen mid-stream (as the receive loop does for target
+            # matching) must not corrupt the committed state.
+            _ = parser.screen
+    return parser.screen
+
+
+# ── crafted byte streams that stress the chunk boundaries ─────────────────────
+
+CRAFTED = [
+    b'',
+    b'hello world',
+    b'\x1b[1;37;44mHELLO\x1b[0m',                       # color strip
+    b'BEFORE\x1b[2JAFTER',                              # clear screen
+    b'A\x1b[2JB\x1b[2JC',                               # only last 2J wins
+    b'helloworld\x1b[1;6H\x1b[K',                       # erase to EOL
+    b'abc \x08def',                                     # single ' \x08'
+    b'a    \x08\x08\x08\x08b',                          # cascading ' \x08'
+    b'   \x08\x08\x08\x08\x08x',                        # more BS than spaces
+    b'\x1b[5;3HHi\nNext',                               # positioning + newline
+    b'  \x08\x08\xe2\x97\x8e tail',                     # ◎ via reservation pair
+    b'\x1b[10;1H  \x08\x08\xe2\x80\xbb\x1b[10;3H \xe7\x99\xbc\xe4\xbf\xa1\xe7\xab\x99: x',
+    b'\x1b[6nHELLO',                                    # unknown escape -> bail
+    b'line1\r\nline2',                                  # CRLF
+    b'helloworld\x1b[1;2HXY',                           # replace mode
+    b'\x1b[100;1HX',                                    # row clamp
+    b'\xe2\x94\x80' * 40,                               # long box-drawing run
+    '中文範例'.encode('utf-8'),                          # wide chars
+]
+
+
+@pytest.mark.parametrize('data', CRAFTED)
+@pytest.mark.parametrize('chunk', CHUNK_SIZES)
+def test_crafted_parity_utf8(data, chunk):
+    ref = VT100Parser(data, 'utf-8', 24).screen
+    assert feed_in_chunks(data, 'utf-8', chunk) == ref
+    assert feed_in_chunks(data, 'utf-8', chunk, read_each=True) == ref
+
+
+BIG5_TEXTS = ['中文', '程式範例', '◎[軟工]※←→─▄▆', '看板《Python》 測試']
+
+
+@pytest.mark.parametrize('text', BIG5_TEXTS)
+@pytest.mark.parametrize('chunk', [1, 2, 3, 5, 64])
+def test_big5uao_parity(text, chunk):
+    data = text.encode('big5')
+    ref = VT100Parser(data, 'big5uao', 24).screen
+    assert feed_in_chunks(data, 'big5uao', chunk) == ref
+
+
+def test_big5uao_realistic_stream_parity():
+    # A PTT-shaped big5uao screen: clear, positioning, wide chars via the
+    # ' \x08' reservation pattern, box drawing, erase-to-EOL, comment glyphs.
+    def b5(s):
+        return s.encode('big5')
+
+    stream = (
+        b'\x1b[2J\x1b[1;1H' + b5('看板《Python》')
+        + b'\x1b[3;1H  \x08\x08' + b5('◎') + b5(' [問題] 測試中文標題 ')
+        + b'\x1b[5;1H' + b5('※ 發信站: 批踢踢實業坊(ptt.cc)')
+        + b'\x1b[6;1H' + b5('推 user123: 這是一則推文') + b'\x1b[K'
+        + b'\x1b[10;1H' + b5('─' * 30)
+        + b'\x1b[23;1H' + b5('瀏覽 第 1/2 頁')
+    )
+    ref = VT100Parser(stream, 'big5uao', 24).screen
+    for chunk in CHUNK_SIZES:
+        assert feed_in_chunks(stream, 'big5uao', chunk) == ref
+        assert feed_in_chunks(stream, 'big5uao', chunk, read_each=True) == ref
+
+
+# ── real captured PTT byte streams, every boundary ────────────────────────────
+
+FIXTURES = sorted(glob.glob(str(FIXTURE_DIR / '*.bin')))
+
+
+@pytest.mark.parametrize('fixture', FIXTURES, ids=lambda p: Path(p).stem)
+@pytest.mark.parametrize('chunk', CHUNK_SIZES)
+def test_fixture_parity(fixture, chunk):
+    data = Path(fixture).read_bytes()
+    encoding = Path(fixture[:-4] + '.encoding').read_text().strip()
+    ref = VT100Parser(data, encoding, 24).screen
+    assert feed_in_chunks(data, encoding, chunk) == ref
+    assert feed_in_chunks(data, encoding, chunk, read_each=True) == ref
+
+
+@pytest.mark.parametrize('fixture', FIXTURES, ids=lambda p: Path(p).stem)
+def test_fixture_parity_random_chunking(fixture):
+    # Randomised chunk boundaries with a mid-stream .screen read each step —
+    # exactly how the receive loop drives the parser.
+    data = Path(fixture).read_bytes()
+    encoding = Path(fixture[:-4] + '.encoding').read_text().strip()
+    ref = VT100Parser(data, encoding, 24).screen
+    for seed in range(25):
+        rnd = random.Random(seed)
+        parser = IncrementalScreen(encoding, 24)
+        i = 0
+        while i < len(data):
+            n = rnd.randint(1, 9)
+            parser.feed(data[i:i + n])
+            _ = parser.screen
+            i += n
+        assert parser.screen == ref, f'seed={seed}'

pyptt-2.2.6/tests/test_set_signature_file.py

@@ -0,0 +1,32 @@
+"""Integration test for set_signature_file (the PTT "名片檔"/plan).
+
+Self-restoring round-trip: capture the account's current signature file via
+`get_user`, overwrite it with an identifiable marker, verify the marker shows
+up on a fresh `get_user` read, then set it back to the original content so
+the test doesn't leave a permanent change on the account.
+"""
+import time
+
+from PyPtt import UserField
+
+
+def test_set_signature_file_round_trip(ptt_bots):
+    for ptt_bot in ptt_bots:
+        original = ptt_bot.get_user(ptt_bot.ptt_id)[UserField.signature_file]
+
+        marker = f'PyPtt set_signature_file test {int(time.time())}'
+        try:
+            ptt_bot.set_signature_file(marker)
+            time.sleep(1)
+
+            updated = ptt_bot.get_user(ptt_bot.ptt_id)[UserField.signature_file]
+            assert marker in updated.strip()
+        finally:
+            # Restore -- must always run, even if the assertion above fails.
+            ptt_bot.set_signature_file(original)
+            time.sleep(1)
+
+            restored = ptt_bot.get_user(ptt_bot.ptt_id)[UserField.signature_file]
+            # Tolerate trailing whitespace/blank-line drift from the
+            # editor round-trip; compare on stripped content.
+            assert restored.strip() == original.strip()