alexion/dev
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

chore(vendor): import pi 1.3.0

Browse Source 
Imported as a linear vendor commit so forge rebases do not replay the upstream root tree.

git-subtree-dir: third_party/pi

git-subtree-split: 71023aec1546d6d677a65669c5665c0927ce0dd0
This commit is contained in:
alexion
2026-10-15 17:58:48 -04:00
parent 719370c8f7
commit 6081e3e325
1743 changed files with 205460 additions and 131705 deletions
Download Patch File Download Diff File Expand all files Collapse all files

171
third_party/pi/packages/tui/src/utils.ts vendored
View File

@@ -54,6 +54,14 @@ const widthCache = new Map<string, number>();
export const cjkBreakRegex =
/[\p{Script_Extensions=Han}\p{Script_Extensions=Hiragana}\p{Script_Extensions=Katakana}\p{Script_Extensions=Hangul}\p{Script_Extensions=Bopomofo}]/u;
// CJK letters remain part of words and paths; only punctuation can separate prose from completions.
export const cjkPunctuationRegex = new RegExp(
`(?:(?=\\p{Punctuation})${cjkBreakRegex.source}|[，．：；！？（）［］｛｝“”‘’…—])`,
"u",
);
export const autocompleteSeparatorRegex = new RegExp(`(?:\\s|${cjkPunctuationRegex.source})`, "u");
export const autocompleteBoundaryRegex = new RegExp(`(?:^|${autocompleteSeparatorRegex.source})`, "u");
function isPrintableAscii(str: string): boolean {
for (let i = 0; i < str.length; i++) {
const code = str.charCodeAt(i);
@@ -172,8 +180,10 @@ function finalizeTruncatedResult(
* check to avoid running the RGI_Emoji regex unnecessarily.
*/
function graphemeWidth(segment: string): number {
if (segment === "\t") {
return 3;
if (segment.length === 1) {
const code = segment.charCodeAt(0);
if (code >= 0x20 && code <= 0x7e) return 1;
if (code === 0x09) return 3;
}
// Some marks occupy cells even without a base character.
@@ -242,9 +252,11 @@ export function visibleWidth(str: string): number {
return 0;
}
// Fast path: pure ASCII printable
if (isPrintableAscii(str)) {
return str.length;
// Fast path: printable ASCII, tabs, and ANSI escape sequences. Styled lines take this path, so
// re-rendering after a theme change does not run grapheme segmentation on every line.
const asciiWidth = asciiVisibleWidth(str);
if (asciiWidth !== -1) {
return asciiWidth;
}
// Check cache
@@ -258,22 +270,25 @@ export function visibleWidth(str: string): number {
if (str.includes("\t")) {
clean = clean.replace(/\t/g, " ");
}
if (clean.includes("\x1b")) {
let escapeIndex = clean.indexOf("\x1b");
if (escapeIndex !== -1) {
// Strip supported ANSI/OSC/APC escape sequences in one pass.
// This covers CSI styling/cursor codes, OSC hyperlinks and prompt markers,
// and APC sequences like CURSOR_MARKER.
let stripped = "";
let i = 0;
while (i < clean.length) {
const ansi = extractAnsiCode(clean, i);
if (ansi) {
i += ansi.length;
continue;
let copyFrom = 0;
while (escapeIndex !== -1) {
const length = ansiCodeLength(clean, escapeIndex);
if (length > 0) {
stripped += clean.slice(copyFrom, escapeIndex);
escapeIndex += length;
copyFrom = escapeIndex;
} else {
escapeIndex++;
}
stripped += clean[i];
i++;
escapeIndex = clean.indexOf("\x1b", escapeIndex);
}
clean = stripped;
clean = stripped + clean.slice(copyFrom);
}
// Calculate width
@@ -404,43 +419,66 @@ export function normalizeTerminalOutput(str: string): string {
* Extract ANSI escape sequences from a string at the given position.
*/
export function extractAnsiCode(str: string, pos: number): { code: string; length: number } | null {
if (pos >= str.length || str[pos] !== "\x1b") return null;
const length = ansiCodeLength(str, pos);
return length > 0 ? { code: str.substring(pos, pos + length), length } : null;
}
/**
* Width of a string made of printable ASCII, tabs, and ANSI escape sequences, or -1 if it contains
* anything else. Matches `visibleWidth` for those strings without allocating.
*/
function asciiVisibleWidth(str: string): number {
let width = 0;
let i = 0;
while (i < str.length) {
const code = str.charCodeAt(i);
if (code >= 0x20 && code <= 0x7e) {
width++;
i++;
} else if (code === 0x09) {
width += 3;
i++;
} else if (code === 0x1b) {
const length = ansiCodeLength(str, i);
if (length === 0) return -1;
i += length;
} else {
return -1;
}
}
return width;
}
/** Length of the ANSI/OSC/APC escape sequence starting at `pos`, or 0 if there is none. */
function ansiCodeLength(str: string, pos: number): number {
if (pos >= str.length || str.charCodeAt(pos) !== 0x1b) return 0;
const next = str[pos + 1];
// CSI sequence: ESC [ ... m/G/K/H/J
if (next === "[") {
let j = pos + 2;
while (j < str.length && !/[mGKHJ]/.test(str[j]!)) j++;
if (j < str.length) return { code: str.substring(pos, j + 1), length: j + 1 - pos };
return null;
for (let j = pos + 2; j < str.length; j++) {
const c = str.charCodeAt(j);
// m, G, K, H, J
if (c === 0x6d || c === 0x47 || c === 0x4b || c === 0x48 || c === 0x4a) return j + 1 - pos;
}
return 0;
}
// OSC sequence: ESC ] ... BEL or ESC ] ... ST (ESC \)
// Used for hyperlinks (OSC 8), window titles, etc.
if (next === "]") {
let j = pos + 2;
while (j < str.length) {
if (str[j] === "\x07") return { code: str.substring(pos, j + 1), length: j + 1 - pos };
if (str[j] === "\x1b" && str[j + 1] === "\\") return { code: str.substring(pos, j + 2), length: j + 2 - pos };
j++;
}
return null;
}
// APC sequence: ESC _ ... BEL or ESC _ ... ST (ESC \)
// Used for cursor marker and application-specific commands
if (next === "_") {
let j = pos + 2;
while (j < str.length) {
if (str[j] === "\x07") return { code: str.substring(pos, j + 1), length: j + 1 - pos };
if (str[j] === "\x1b" && str[j + 1] === "\\") return { code: str.substring(pos, j + 2), length: j + 2 - pos };
j++;
if (next === "]" || next === "_") {
for (let j = pos + 2; j < str.length; j++) {
const c = str.charCodeAt(j);
if (c === 0x07) return j + 1 - pos;
if (c === 0x1b && str[j + 1] === "\\") return j + 2 - pos;
}
return null;
return 0;
}
return null;
return 0;
}
type Osc8Terminator = "\x07" | "\x1b\\";
@@ -731,15 +769,16 @@ class AnsiCodeTracker {
}
function updateTrackerFromText(text: string, tracker: AnsiCodeTracker): void {
let i = 0;
while (i < text.length) {
const ansiResult = extractAnsiCode(text, i);
if (ansiResult) {
tracker.process(ansiResult.code);
i += ansiResult.length;
let i = text.indexOf("\x1b");
while (i !== -1) {
const length = ansiCodeLength(text, i);
if (length > 0) {
tracker.process(text.substring(i, i + length));
i += length;
} else {
i++;
}
i = text.indexOf("\x1b", i);
}
}
@@ -750,6 +789,10 @@ export function getActiveBackgroundAnsi(text: string): string {
return tracker.getActiveBackgroundCode();
}
function* graphemeSegments(text: string): Generator<string> {
for (const { segment } of graphemeSegmenter.segment(text)) yield segment;
}
/**
* Split text into words while keeping ANSI codes attached.
*/
@@ -778,14 +821,19 @@ function splitIntoTokensWithAnsi(text: string): string[] {
continue;
}
let end = i;
while (end < text.length && !extractAnsiCode(text, end)) {
end++;
// Visible text runs up to the next escape sequence.
let end = text.indexOf("\x1b", i + 1);
while (end !== -1 && ansiCodeLength(text, end) === 0) {
end = text.indexOf("\x1b", end + 1);
}
if (end === -1) end = text.length;
for (const { segment } of graphemeSegmenter.segment(text.slice(i, end))) {
const chunk = text.slice(i, end);
// Printable ASCII characters are single graphemes, so skip the segmenter for them.
const ascii = isPrintableAscii(chunk);
for (const segment of ascii ? chunk : graphemeSegments(chunk)) {
const segmentIsSpace = segment === " ";
if (!segmentIsSpace && cjkBreakRegex.test(segment)) {
if (!ascii && !segmentIsSpace && cjkBreakRegex.test(segment)) {
flushCurrent();
const token = pendingAnsi + segment;
pendingAnsi = "";
@@ -840,6 +888,15 @@ function splitIntoTokensWithAnsi(text: string): string[] {
* @param width - Maximum visible width per line
* @returns Array of wrapped lines (NOT padded to width)
*/
/**
* Flatten cached lines. V8 keeps a string built by concatenation as a tree of its parts until something reads it
* whole, and a cached line kept as such a tree retains several times its own size. Converting a string to a number
* reads it whole, so V8 flattens it in place; the strings' values do not change.
*/
export function flattenLines(lines: readonly string[]): void {
for (const line of lines) Number(line);
}
export function wrapTextWithAnsi(text: string, width: number): string[] {
if (!text) {
return [""];
@@ -1225,8 +1282,11 @@ export function sliceWithWidth(
while (i < line.length) {
const ansi = extractAnsiCode(line, i);
if (ansi) {
if (currentCol >= startCol && currentCol < endCol) result += ansi.code;
else if (currentCol < startCol) pendingAnsi += ansi.code;
if (currentCol >= startCol && currentCol < endCol) {
// Keep original order: codes from before the range must precede codes at the boundary
result += pendingAnsi + ansi.code;
pendingAnsi = "";
} else if (currentCol < startCol) pendingAnsi += ansi.code;
i += ansi.length;
continue;
}
@@ -1291,8 +1351,13 @@ export function extractSegments(
// Include ANSI codes in their respective segments
if (currentCol < beforeEnd) {
pendingAnsiBefore += ansi.code;
} else if (currentCol >= afterStart && currentCol < afterEnd && afterStarted) {
// Only include after we've started "after" (styling already prepended)
} else if (
currentCol >= afterStart &&
currentCol < afterEnd &&
(afterStarted || ansi.code.startsWith("\x1b_"))
) {
// SGR codes are only included after "after" started (styling already prepended). Zero-width APC
// markers like CURSOR_MARKER are not tracked styling, so keep them at the boundary too.
after += ansi.code;
}
i += ansi.length;