[3.13] gh-119517: Fixes for pasting in pyrepl (GH-120253) (#120353)

gh-119517: Fixes for pasting in pyrepl (GH-120253)

* Remove pyrepl's optimization for self-insert

This will be replaced by a less specialized optimization.

* Use line-buffering when pyrepl echoes pastes

Previously echoing was totally suppressed until the entire command had
been pasted and the terminal ended paste mode, but this gives the user
no feedback to indicate that an operation is in progress. Drawing
something to the screen once per line strikes a balance between
perceived responsiveness and performance.

* Remove dead code from pyrepl

`msg_at_bottom` is always true.

* Speed up pyrepl's screen rendering computation

The Reader in pyrepl doesn't hold a complete representation of the
screen area being drawn as persistent state. Instead, it recomputes it,
on each keypress. This is fast enough for a few hundred bytes, but
incredibly slow as the input buffer grows into the kilobytes (likely
because of pasting).

Rather than making some expensive and expansive changes to the repl's
internal representation of the screen, add some caching: remember some
data from one refresh to the next about what was drawn to the screen
and, if we don't find anything that has invalidated the results that
were computed last time around, reuse them. To keep this caching as
simple as possible, all we'll do is look for lines in the buffer that
were above the cursor the last time we were asked to update the screen,
and that are still above the cursor now. We assume that nothing can
affect a line that comes before both the old and new cursor location
without us being informed. Based on this assumption, we can reuse old
lines, which drastically speeds up the overwhelmingly common case where
the user is typing near the end of the buffer.

* Speed up pyrepl prompt drawing

Cache the `can_colorize()` call rather than repeatedly recomputing it.
This call looks up an environment variable, and is called once per
character typed at the REPL. The environment variable lookup shows up as
a hot spot when profiling, and we don't expect this to change while the
REPL is running.

* Speed up pasting multiple lines into the REPL

Previously, we were checking whether the command should be accepted each
time a line break was encountered, but that's not the expected behavior.
In bracketed paste mode, we expect everything pasted to be part of
a single block of code, and encountering a newline shouldn't behave like
a user pressing <Enter> to execute a command. The user should always
have a chance to review the pasted command before running it.

* Use a read buffer for input in pyrepl

Previously we were reading one byte at a time, which causes much slower
IO than necessary. Instead, read in chunks, processing previously read
data before asking for more.

* Optimize finding width of a single character

`wlen` finds the width of a multi-character string by adding up the
width of each character, and then subtracting the width of any escape
sequences. It's often called for single character strings, however,
which can't possibly contain escape sequences. Optimize for that case.

* Optimize disp_str for ASCII characters

Since every ASCII character is known to display as single width, we can
avoid not only the Unicode data lookup in `disp_str` but also the one
hidden in `str_width` for them.

* Speed up cursor movements in long pyrepl commands

When the current pyrepl command buffer contains many lines, scrolling up
becomes slow. We have optimizations in place to reuse lines above the
cursor position from one refresh to the next, but don't currently try to
reuse lines below the cursor position in the same way, so we wind up
with quadratic behavior where all lines of the buffer below the cursor
are recomputed each time the cursor moves up another line.

Optimize this by only computing one screen's worth of lines beyond the
cursor position. Any lines beyond that can't possibly be shown by the
console, and bounding this makes scrolling up have linear time
complexity instead.

---------

(cherry picked from commit 32a0faba43)

Signed-off-by: Matt Wozniski <mwozniski@bloomberg.net>
Co-authored-by: Matt Wozniski <mwozniski@bloomberg.net>
Co-authored-by: Pablo Galindo <pablogsal@gmail.com>
This commit is contained in:
Miss Islington (bot) 2024-06-11 19:09:28 +02:00 committed by GitHub
parent 2dd07fd899
commit 4936575017
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
6 changed files with 134 additions and 57 deletions

View File

@ -368,8 +368,6 @@ def do(self) -> None:
r = self.reader
text = self.event * r.get_arg()
r.insert(text)
if len(text) == 1 and r.pos == len(r.buffer):
r.calc_screen = r.append_to_screen
class insert_nl(EditCommand):
@ -483,4 +481,3 @@ def do(self) -> None:
self.reader.paste_mode = False
self.reader.in_bracketed_paste = False
self.reader.dirty = True
self.reader.calc_screen = self.reader.calc_complete_screen

View File

@ -209,10 +209,6 @@ def do(self) -> None:
r = self.reader # type: ignore[assignment]
commands.self_insert.do(self)
if r.cmpltn_menu_visible or r.cmpltn_message_visible:
r.calc_screen = r.calc_complete_screen
if r.cmpltn_menu_visible:
stem = r.get_stem()
if len(stem) < 1:
@ -261,8 +257,8 @@ def after_command(self, cmd: Command) -> None:
if not isinstance(cmd, (complete, self_insert)):
self.cmpltn_reset()
def calc_complete_screen(self) -> list[str]:
screen = super().calc_complete_screen()
def calc_screen(self) -> list[str]:
screen = super().calc_screen()
if self.cmpltn_menu_visible:
ly = self.lxy[1]
screen[ly:ly] = self.cmpltn_menu

View File

@ -35,15 +35,13 @@
# types
Command = commands.Command
if False:
from typing import Callable
from .types import Callback, SimpleContextManager, KeySpec, CommandName
CalcScreen = Callable[[], list[str]]
def disp_str(buffer: str) -> tuple[str, list[int]]:
"""disp_str(buffer:string) -> (string, [int])
Return the string that should be the printed represenation of
Return the string that should be the printed representation of
|buffer| and a list detailing where the characters of |buffer|
get used up. E.g.:
@ -54,11 +52,17 @@ def disp_str(buffer: str) -> tuple[str, list[int]]:
b: list[int] = []
s: list[str] = []
for c in buffer:
if ord(c) > 128 and unicodedata.category(c).startswith("C"):
if ord(c) < 128:
s.append(c)
b.append(1)
elif unicodedata.category(c).startswith("C"):
c = r"\u%04x" % ord(c)
s.append(c)
b.append(wlen(c))
b.extend([0] * (len(c) - 1))
s.append(c)
b.append(str_width(c))
b.extend([0] * (len(c) - 1))
else:
s.append(c)
b.append(str_width(c))
return "".join(s), b
@ -230,7 +234,6 @@ class Reader:
commands: dict[str, type[Command]] = field(default_factory=make_default_commands)
last_command: type[Command] | None = None
syntax_table: dict[str, int] = field(default_factory=make_default_syntax_table)
msg_at_bottom: bool = True
keymap: tuple[tuple[str, str], ...] = ()
input_trans: input.KeymapTranslator = field(init=False)
input_trans_stack: list[input.KeymapTranslator] = field(default_factory=list)
@ -238,8 +241,52 @@ class Reader:
screeninfo: list[tuple[int, list[int]]] = field(init=False)
cxy: tuple[int, int] = field(init=False)
lxy: tuple[int, int] = field(init=False)
calc_screen: CalcScreen = field(init=False)
scheduled_commands: list[str] = field(default_factory=list)
can_colorize: bool = False
## cached metadata to speed up screen refreshes
@dataclass
class RefreshCache:
in_bracketed_paste: bool = False
screen: list[str] = field(default_factory=list)
screeninfo: list[tuple[int, list[int]]] = field(init=False)
line_end_offsets: list[int] = field(default_factory=list)
pos: int = field(init=False)
cxy: tuple[int, int] = field(init=False)
dimensions: tuple[int, int] = field(init=False)
def update_cache(self,
reader: Reader,
screen: list[str],
screeninfo: list[tuple[int, list[int]]],
) -> None:
self.in_bracketed_paste = reader.in_bracketed_paste
self.screen = screen.copy()
self.screeninfo = screeninfo.copy()
self.pos = reader.pos
self.cxy = reader.cxy
self.dimensions = reader.console.width, reader.console.height
def valid(self, reader: Reader) -> bool:
dimensions = reader.console.width, reader.console.height
dimensions_changed = dimensions != self.dimensions
paste_changed = reader.in_bracketed_paste != self.in_bracketed_paste
return not (dimensions_changed or paste_changed)
def get_cached_location(self, reader: Reader) -> tuple[int, int]:
offset = 0
earliest_common_pos = min(reader.pos, self.pos)
num_common_lines = len(self.line_end_offsets)
while num_common_lines > 0:
offset = self.line_end_offsets[num_common_lines - 1]
if earliest_common_pos > offset:
break
num_common_lines -= 1
else:
offset = 0
return offset, num_common_lines
last_refresh_cache: RefreshCache = field(default_factory=RefreshCache)
def __post_init__(self) -> None:
# Enable the use of `insert` without a `prepare` call - necessary to
@ -252,53 +299,60 @@ def __post_init__(self) -> None:
self.screeninfo = [(0, [])]
self.cxy = self.pos2xy()
self.lxy = (self.pos, 0)
self.calc_screen = self.calc_complete_screen
self.can_colorize = can_colorize()
self.last_refresh_cache.screeninfo = self.screeninfo
self.last_refresh_cache.pos = self.pos
self.last_refresh_cache.cxy = self.cxy
self.last_refresh_cache.dimensions = (0, 0)
def collect_keymap(self) -> tuple[tuple[KeySpec, CommandName], ...]:
return default_keymap
def append_to_screen(self) -> list[str]:
new_screen = self.screen.copy() or ['']
def calc_screen(self) -> list[str]:
"""Translate changes in self.buffer into changes in self.console.screen."""
# Since the last call to calc_screen:
# screen and screeninfo may differ due to a completion menu being shown
# pos and cxy may differ due to edits, cursor movements, or completion menus
new_character = self.buffer[-1]
new_character_len = wlen(new_character)
# Lines that are above both the old and new cursor position can't have changed,
# unless the terminal has been resized (which might cause reflowing) or we've
# entered or left paste mode (which changes prompts, causing reflowing).
num_common_lines = 0
offset = 0
if self.last_refresh_cache.valid(self):
offset, num_common_lines = self.last_refresh_cache.get_cached_location(self)
last_line_len = wlen(new_screen[-1])
if last_line_len + new_character_len >= self.console.width: # We need to wrap here
new_screen[-1] += '\\'
self.screeninfo[-1][1].append(1)
new_screen.append(self.buffer[-1])
self.screeninfo.append((0, [new_character_len]))
else:
new_screen[-1] += self.buffer[-1]
self.screeninfo[-1][1].append(new_character_len)
self.cxy = self.pos2xy()
screen = self.last_refresh_cache.screen
del screen[num_common_lines:]
# Reset the function that is used for completing the screen
self.calc_screen = self.calc_complete_screen
return new_screen
screeninfo = self.last_refresh_cache.screeninfo
del screeninfo[num_common_lines:]
last_refresh_line_end_offsets = self.last_refresh_cache.line_end_offsets
del last_refresh_line_end_offsets[num_common_lines:]
def calc_complete_screen(self) -> list[str]:
"""The purpose of this method is to translate changes in
self.buffer into changes in self.screen. Currently it rips
everything down and starts from scratch, which whilst not
especially efficient is certainly simple(r).
"""
lines = self.get_unicode().split("\n")
screen: list[str] = []
screeninfo: list[tuple[int, list[int]]] = []
pos = self.pos
for ln, line in enumerate(lines):
pos -= offset
lines = "".join(self.buffer[offset:]).split("\n")
cursor_found = False
lines_beyond_cursor = 0
for ln, line in enumerate(lines, num_common_lines):
ll = len(line)
if 0 <= pos <= ll:
if self.msg and not self.msg_at_bottom:
for mline in self.msg.split("\n"):
screen.append(mline)
screeninfo.append((0, []))
self.lxy = pos, ln
cursor_found = True
elif cursor_found:
lines_beyond_cursor += 1
if lines_beyond_cursor > self.console.height:
# No need to keep formatting lines.
# The console can't show them.
break
prompt = self.get_prompt(ln, ll >= pos >= 0)
while "\n" in prompt:
pre_prompt, _, prompt = prompt.partition("\n")
last_refresh_line_end_offsets.append(offset)
screen.append(pre_prompt)
screeninfo.append((0, []))
pos -= ll + 1
@ -306,6 +360,8 @@ def calc_complete_screen(self) -> list[str]:
l, l2 = disp_str(line)
wrapcount = (wlen(l) + lp) // self.console.width
if wrapcount == 0:
offset += ll + 1 # Takes all of the line plus the newline
last_refresh_line_end_offsets.append(offset)
screen.append(prompt + l)
screeninfo.append((lp, l2))
else:
@ -321,11 +377,14 @@ def calc_complete_screen(self) -> list[str]:
column += character_width
pre = prompt if i == 0 else ""
if len(l) > index_to_wrap_before:
offset += index_to_wrap_before
post = "\\"
after = [1]
else:
offset += index_to_wrap_before + 1 # Takes the newline
post = ""
after = []
last_refresh_line_end_offsets.append(offset)
screen.append(pre + l[:index_to_wrap_before] + post)
screeninfo.append((prelen, l2[:index_to_wrap_before] + after))
l = l[index_to_wrap_before:]
@ -333,10 +392,12 @@ def calc_complete_screen(self) -> list[str]:
i += 1
self.screeninfo = screeninfo
self.cxy = self.pos2xy()
if self.msg and self.msg_at_bottom:
if self.msg:
for mline in self.msg.split("\n"):
screen.append(mline)
screeninfo.append((0, []))
self.last_refresh_cache.update_cache(self, screen, screeninfo)
return screen
@staticmethod
@ -456,7 +517,7 @@ def get_prompt(self, lineno: int, cursor_on_line: bool) -> str:
`lineno'."""
if self.arg is not None and cursor_on_line:
prompt = f"(arg: {self.arg}) "
elif self.paste_mode:
elif self.paste_mode and not self.in_bracketed_paste:
prompt = "(paste) "
elif "\n" in self.buffer:
if lineno == 0:
@ -468,7 +529,7 @@ def get_prompt(self, lineno: int, cursor_on_line: bool) -> str:
else:
prompt = self.ps1
if can_colorize():
if self.can_colorize:
prompt = f"{ANSIColors.BOLD_MAGENTA}{prompt}{ANSIColors.RESET}"
return prompt
@ -604,6 +665,9 @@ def update_screen(self) -> None:
def refresh(self) -> None:
"""Recalculate and refresh the screen."""
if self.in_bracketed_paste and self.buffer and not self.buffer[-1] == "\n":
return
# this call sets up self.cxy, so call it first.
self.screen = self.calc_screen()
self.console.refresh(self.screen, self.cxy)
@ -627,7 +691,7 @@ def do_cmd(self, cmd: tuple[str, list[str]]) -> None:
self.after_command(command)
if self.dirty and not self.in_bracketed_paste:
if self.dirty:
self.refresh()
else:
self.update_cursor()

View File

@ -263,6 +263,10 @@ def do(self) -> None:
r = self.reader # type: ignore[assignment]
r.dirty = True # this is needed to hide the completion menu, if visible
if self.reader.in_bracketed_paste:
r.insert("\n")
return
# if there are already several lines and the cursor
# is not on the last one, always insert a new \n.
text = r.get_unicode()

View File

@ -150,6 +150,8 @@ def __init__(
self.pollob = poll()
self.pollob.register(self.input_fd, select.POLLIN)
self.input_buffer = b""
self.input_buffer_pos = 0
curses.setupterm(term or None, self.output_fd)
self.term = term
@ -197,6 +199,18 @@ def _my_getstr(cap: str, optional: bool = False) -> bytes | None:
self.event_queue = EventQueue(self.input_fd, self.encoding)
self.cursor_visible = 1
def __read(self, n: int) -> bytes:
if not self.input_buffer or self.input_buffer_pos >= len(self.input_buffer):
self.input_buffer = os.read(self.input_fd, 10000)
ret = self.input_buffer[self.input_buffer_pos : self.input_buffer_pos + n]
self.input_buffer_pos += len(ret)
if self.input_buffer_pos >= len(self.input_buffer):
self.input_buffer = b""
self.input_buffer_pos = 0
return ret
def change_encoding(self, encoding: str) -> None:
"""
Change the encoding used for I/O operations.
@ -373,7 +387,7 @@ def get_event(self, block: bool = True) -> Event | None:
while self.event_queue.empty():
while True:
try:
self.push_char(os.read(self.input_fd, 1))
self.push_char(self.__read(1))
except OSError as err:
if err.errno == errno.EINTR:
if not self.event_queue.empty():
@ -491,7 +505,7 @@ def getpending(self):
e.raw += e.raw
amount = struct.unpack("i", ioctl(self.input_fd, FIONREAD, b"\0\0\0\0"))[0]
raw = os.read(self.input_fd, amount)
raw = self.__read(amount)
data = str(raw, self.encoding, "replace")
e.data += data
e.raw += raw
@ -514,7 +528,7 @@ def getpending(self):
e.raw += e.raw
amount = 10000
raw = os.read(self.input_fd, amount)
raw = self.__read(amount)
data = str(raw, self.encoding, "replace")
e.data += data
e.raw += raw

View File

@ -16,6 +16,8 @@ def str_width(c: str) -> int:
def wlen(s: str) -> int:
if len(s) == 1:
return str_width(s)
length = sum(str_width(i) for i in s)
# remove lengths of any escape sequences
sequence = ANSI_ESCAPE_SEQUENCE.findall(s)