@oh-my-pi/pi-coding-agent 15.5.3 → 15.5.4

package/src/hashline/tokenizer.ts

@@ -1,473 +0,0 @@
-import { ABORT_MARKER, BEGIN_PATCH_MARKER, END_PATCH_MARKER } from "./constants";
-import {
-	describeAnchorExamples,
-	HL_FILE_HASH_SEP,
-	HL_FILE_PREFIX,
-	HL_OP_DELETE,
-	HL_OP_INSERT_AFTER,
-	HL_OP_INSERT_BEFORE,
-	HL_OP_REPLACE,
-	HL_PAYLOAD_PREFIX,
-} from "./hash";
-import type { Anchor, HashlineCursor } from "./types";
-
-const CHAR_LINE_FEED = 10;
-const CHAR_CARRIAGE_RETURN = 13;
-const CHAR_ZERO = 48;
-const CHAR_NINE = 57;
-const CHAR_HASH = 35;
-const CHAR_TAB = 9;
-const CHAR_SPACE = 32;
-const CHAR_LOWER_A = 97;
-const CHAR_LOWER_F = 102;
-const CHAR_PILCROW = HL_FILE_PREFIX.charCodeAt(0);
-const CHAR_PAYLOAD_PREFIX = HL_PAYLOAD_PREFIX.charCodeAt(0);
-const FILE_HASH_LENGTH = 4;
-
-function isDigitCode(code: number): boolean {
-	return code >= CHAR_ZERO && code <= CHAR_NINE;
-}
-
-function isNonZeroDigitCode(code: number): boolean {
-	return code > CHAR_ZERO && code <= CHAR_NINE;
-}
-
-function isDecorationCode(code: number): boolean {
-	return code === 42 || code === 45 || code === 62;
-}
-
-function isHexDigitCode(code: number): boolean {
-	return isDigitCode(code) || (code >= CHAR_LOWER_A && code <= CHAR_LOWER_F);
-}
-
-function skipWhitespace(line: string, index: number, end = line.length): number {
-	return end - line.slice(index, end).trimStart().length;
-}
-
-function trimEndIndex(line: string): number {
-	return line.trimEnd().length;
-}
-
-function isEmptyLine(line: string): boolean {
-	return line.length === 0;
-}
-
-function markerLineEquals(line: string, marker: string): boolean {
-	return line.trimEnd() === marker;
-}
-
-/**
- * Split a hashline diff into individual lines without losing the trailing
- * empty line that callers may rely on for explicit blank payloads. CRLF pairs
- * are normalized to a single line break.
- *
- * This mirrors the line-splitting performed by {@link HashlineTokenizer}'s
- * streaming drain loop and is kept for non-streaming callers that prefer
- * a single-shot split.
- */
-export function splitHashlineLines(text: string): string[] {
-	if (text.length === 0) return [""];
-
-	const lines: string[] = [];
-	let start = 0;
-	for (let index = 0; index < text.length; index++) {
-		if (text.charCodeAt(index) !== CHAR_LINE_FEED) continue;
-		let end = index;
-		if (end > start && text.charCodeAt(end - 1) === CHAR_CARRIAGE_RETURN) end--;
-		lines.push(text.slice(start, end));
-		start = index + 1;
-	}
-
-	if (start < text.length) {
-		let end = text.length;
-		if (end > start && text.charCodeAt(end - 1) === CHAR_CARRIAGE_RETURN) end--;
-		lines.push(text.slice(start, end));
-	}
-	return lines;
-}
-
-export function cloneCursor(cursor: HashlineCursor): HashlineCursor {
-	if (cursor.kind === "before_anchor") return { kind: "before_anchor", anchor: { ...cursor.anchor } };
-	if (cursor.kind === "after_anchor") return { kind: "after_anchor", anchor: { ...cursor.anchor } };
-	return cursor;
-}
-
-// Leniently accept anchors copied from read/search output:
-//   - optional leading line-marker decoration (`*`, `>`, `-`)
-//   - the required bare line number
-function skipDecoratedAnchorPrefix(line: string, end = trimEndIndex(line)): number {
-	let index = skipWhitespace(line, 0, end);
-	while (index < end && isDecorationCode(line.charCodeAt(index))) index++;
-	return skipWhitespace(line, index, end);
-}
-
-interface NumberScan {
-	line: number;
-	nextIndex: number;
-}
-
-function scanLineNumber(line: string, index: number, end: number): NumberScan | null {
-	if (index >= end || !isNonZeroDigitCode(line.charCodeAt(index))) return null;
-
-	let lineNumber = 0;
-	let nextIndex = index;
-	while (nextIndex < end) {
-		const code = line.charCodeAt(nextIndex);
-		if (!isDigitCode(code)) break;
-		lineNumber = lineNumber * 10 + (code - CHAR_ZERO);
-		nextIndex++;
-	}
-	return { line: lineNumber, nextIndex };
-}
-
-/** Parse a bare line-number anchor (used by insert ops). Throws on malformed input. */
-export function parseLid(raw: string, lineNum: number): Anchor {
-	const end = trimEndIndex(raw);
-	const numberStart = skipDecoratedAnchorPrefix(raw, end);
-	const number = scanLineNumber(raw, numberStart, end);
-	if (number === null || skipWhitespace(raw, number.nextIndex, end) !== end) {
-		throw new Error(
-			`line ${lineNum}: expected a line number such as ${describeAnchorExamples("119")}; ` +
-				`got ${JSON.stringify(raw)}. Use ${HL_FILE_PREFIX}PATH${HL_FILE_HASH_SEP}hash from your latest read for file-version binding.`,
-		);
-	}
-	return { line: number.line };
-}
-
-export interface ParsedRange {
-	start: Anchor;
-	end: Anchor;
-}
-
-interface RangeScan {
-	range: ParsedRange;
-	nextIndex: number;
-}
-
-function scanRange(line: string, end = trimEndIndex(line)): RangeScan | null {
-	const numberStart = skipDecoratedAnchorPrefix(line, end);
-	const start = scanLineNumber(line, numberStart, end);
-	if (start === null) return null;
-
-	let nextIndex = start.nextIndex;
-	let rangeEnd = start.line;
-	if (nextIndex < end && line.charCodeAt(nextIndex) === 45) {
-		const endNumber = scanLineNumber(line, nextIndex + 1, end);
-		if (endNumber === null) return null;
-		rangeEnd = endNumber.line;
-		nextIndex = endNumber.nextIndex;
-	}
-
-	return {
-		range: { start: { line: start.line }, end: { line: rangeEnd } },
-		nextIndex: skipWhitespace(line, nextIndex, end),
-	};
-}
-
-function startsWithWord(line: string, index: number, end: number, word: string): boolean {
-	if (index + word.length > end) return false;
-	for (let offset = 0; offset < word.length; offset++) {
-		if (line.charCodeAt(index + offset) !== word.charCodeAt(offset)) return false;
-	}
-	return true;
-}
-
-function parseInsertTarget(raw: string, lineNum: number, kind: "before" | "after"): HashlineCursor {
-	const end = trimEndIndex(raw);
-	const targetStart = skipDecoratedAnchorPrefix(raw, end);
-
-	if (startsWithWord(raw, targetStart, end, "BOF") && skipWhitespace(raw, targetStart + 3, end) === end) {
-		return { kind: "bof" };
-	}
-	if (startsWithWord(raw, targetStart, end, "EOF") && skipWhitespace(raw, targetStart + 3, end) === end) {
-		return { kind: "eof" };
-	}
-
-	const cursorKind = kind === "before" ? "before_anchor" : "after_anchor";
-	return { kind: cursorKind, anchor: parseLid(raw, lineNum) };
-}
-
-function scanInlineBody(line: string, index: number): string | undefined {
-	const end = trimEndIndex(line);
-	return index < end ? line.slice(index, end) : undefined;
-}
-
-interface ParsedInsertOp {
-	kind: "insert";
-	cursor: HashlineCursor;
-	inlineBody: string | undefined;
-}
-
-interface ParsedReplaceOp {
-	kind: "replace";
-	range: ParsedRange;
-	inlineBody: string | undefined;
-}
-
-interface ParsedDeleteOp {
-	kind: "delete";
-	range: ParsedRange;
-	trailingPayload: boolean;
-}
-
-type ParsedOp = ParsedInsertOp | ParsedReplaceOp | ParsedDeleteOp;
-
-function tryParseInsertOp(line: string, sigil: string, kind: "before" | "after"): ParsedInsertOp | null {
-	const end = trimEndIndex(line);
-	const targetStart = skipDecoratedAnchorPrefix(line, end);
-
-	let targetEnd: number;
-	if (startsWithWord(line, targetStart, end, "BOF") || startsWithWord(line, targetStart, end, "EOF")) {
-		targetEnd = targetStart + 3;
-	} else {
-		const anchor = scanLineNumber(line, targetStart, end);
-		if (anchor === null) return null;
-		targetEnd = anchor.nextIndex;
-	}
-
-	const opIndex = skipWhitespace(line, targetEnd, end);
-	if (opIndex >= end || line[opIndex] !== sigil) return null;
-
-	// parseInsertTarget can only throw on inputs that already passed the
-	// BOF/EOF/line-number scan above, but guard the throw anyway — the
-	// tokenizer contract forbids it and a future refactor of the prefix
-	// scan must not silently start raising here.
-	try {
-		return {
-			kind: "insert",
-			cursor: parseInsertTarget(line.slice(0, opIndex), 0, kind),
-			inlineBody: scanInlineBody(line, opIndex + sigil.length),
-		};
-	} catch {
-		return null;
-	}
-}
-
-function tryParseReplaceOp(line: string): ParsedReplaceOp | null {
-	const end = trimEndIndex(line);
-	const range = scanRange(line, end);
-	if (range === null || range.nextIndex >= end || line[range.nextIndex] !== HL_OP_REPLACE) return null;
-	return {
-		kind: "replace",
-		range: range.range,
-		inlineBody: scanInlineBody(line, range.nextIndex + HL_OP_REPLACE.length),
-	};
-}
-
-function tryParseDeleteOp(line: string): ParsedDeleteOp | null {
-	const end = trimEndIndex(line);
-	const range = scanRange(line, end);
-	if (range === null || range.nextIndex >= end || line[range.nextIndex] !== HL_OP_DELETE) return null;
-	const afterSigil = range.nextIndex + HL_OP_DELETE.length;
-	return { kind: "delete", range: range.range, trailingPayload: afterSigil !== end };
-}
-
-function tryParseOp(line: string): ParsedOp | null {
-	return (
-		tryParseInsertOp(line, HL_OP_INSERT_BEFORE, "before") ??
-		tryParseInsertOp(line, HL_OP_INSERT_AFTER, "after") ??
-		tryParseReplaceOp(line) ??
-		tryParseDeleteOp(line)
-	);
-}
-
-/**
- * Strict header scan: `¶+` prefix, optional whitespace, path body that excludes
- * whitespace, `#`, and `¶`, optional `#[0-9a-f]{4}` hash suffix, optional
- * trailing whitespace. Returns `null` when any byte deviates from the shape.
- */
-function tryParseHeader(line: string): { path: string; fileHash?: string } | null {
-	const end = trimEndIndex(line);
-	if (end === 0 || line.charCodeAt(0) !== CHAR_PILCROW) return null;
-
-	let index = 0;
-	while (index < end && line.charCodeAt(index) === CHAR_PILCROW) index++;
-	index = skipWhitespace(line, index, end);
-	if (index >= end) return null;
-
-	const pathStart = index;
-	while (index < end) {
-		const code = line.charCodeAt(index);
-		if (code === CHAR_HASH || code === CHAR_PILCROW || code === CHAR_SPACE || code === CHAR_TAB) break;
-		index++;
-	}
-	if (index === pathStart) return null;
-	const path = line.slice(pathStart, index);
-
-	let fileHash: string | undefined;
-	if (index < end && line.charCodeAt(index) === CHAR_HASH) {
-		const hashStart = index + 1;
-		const hashEnd = hashStart + FILE_HASH_LENGTH;
-		if (hashEnd > end) return null;
-		for (let probe = hashStart; probe < hashEnd; probe++) {
-			if (!isHexDigitCode(line.charCodeAt(probe))) return null;
-		}
-		fileHash = line.slice(hashStart, hashEnd);
-		index = hashEnd;
-	}
-
-	// Anything other than trailing whitespace disqualifies the header.
-	if (skipWhitespace(line, index, end) !== end) return null;
-
-	return fileHash !== undefined ? { path, fileHash } : { path };
-}
-
-/**
- * Returns true when the line scans as `LINE!payload` (delete sigil followed by
- * additional content). The executor uses this for the dedicated "deletes only"
- * diagnostic, separate from the standard "unrecognized op" path.
- */
-export function isDeleteOpWithPayload(line: string): boolean {
-	const range = scanRange(line, line.length);
-	return (
-		range !== null &&
-		range.nextIndex < line.length &&
-		line[range.nextIndex] === HL_OP_DELETE &&
-		range.nextIndex + HL_OP_DELETE.length < line.length
-	);
-}
-
-interface TokenBase {
-	/** 1-indexed line number in the original input stream. */
-	lineNum: number;
-}
-
-export type HashlineToken =
-	| (TokenBase & { kind: "blank" })
-	| (TokenBase & { kind: "envelope-begin" })
-	| (TokenBase & { kind: "envelope-end" })
-	| (TokenBase & { kind: "abort" })
-	| (TokenBase & { kind: "header"; path: string; fileHash?: string })
-	| (TokenBase & { kind: "op-insert"; cursor: HashlineCursor; inlineBody: string | undefined })
-	| (TokenBase & { kind: "op-replace"; range: ParsedRange; inlineBody: string | undefined })
-	| (TokenBase & { kind: "op-delete"; range: ParsedRange; trailingPayload: boolean })
-	| (TokenBase & { kind: "payload"; text: string })
-	| (TokenBase & { kind: "raw"; text: string });
-
-function classifyLine(line: string, lineNum: number): HashlineToken {
-	if (isEmptyLine(line)) return { kind: "blank", lineNum };
-	if (markerLineEquals(line, BEGIN_PATCH_MARKER)) return { kind: "envelope-begin", lineNum };
-	if (markerLineEquals(line, END_PATCH_MARKER)) return { kind: "envelope-end", lineNum };
-	if (markerLineEquals(line, ABORT_MARKER)) return { kind: "abort", lineNum };
-
-	if (line.charCodeAt(0) === CHAR_PILCROW) {
-		const header = tryParseHeader(line);
-		if (header !== null) {
-			return header.fileHash !== undefined
-				? { kind: "header", lineNum, path: header.path, fileHash: header.fileHash }
-				: { kind: "header", lineNum, path: header.path };
-		}
-	}
-
-	if (line.charCodeAt(0) === CHAR_PAYLOAD_PREFIX) {
-		return { kind: "payload", lineNum, text: line.slice(HL_PAYLOAD_PREFIX.length) };
-	}
-	const op = tryParseOp(line);
-	if (op !== null) {
-		if (op.kind === "insert") {
-			return { kind: "op-insert", lineNum, cursor: op.cursor, inlineBody: op.inlineBody };
-		}
-		if (op.kind === "replace") {
-			return { kind: "op-replace", lineNum, range: op.range, inlineBody: op.inlineBody };
-		}
-		return { kind: "op-delete", lineNum, range: op.range, trailingPayload: op.trailingPayload };
-	}
-
-	return { kind: "raw", lineNum, text: line };
-}
-
-/**
- * Stateful, line-oriented classifier for hashline diff text. Use the streaming
- * {@link feed}/{@link end} pair to ingest text in chunks (each completed line
- * emits exactly one token; a trailing partial line stays buffered until the
- * next chunk or {@link end}). Use the stateless {@link tokenize}/predicate
- * methods for callers that already hold whole lines and only need
- * classification without buffering.
- */
-export class HashlineTokenizer {
-	#buffer = "";
-	#nextLineNum = 1;
-	#closed = false;
-
-	/**
-	 * Ingest a chunk of input text. Each newline-terminated line in the
-	 * combined buffer produces one token. A trailing partial line (no `\n`
-	 * yet, possibly ending in a lone `\r`) stays buffered until the next
-	 * `feed`/`end` call so CRLF pairs that straddle chunk boundaries are
-	 * still normalized correctly.
-	 */
-	feed(chunk: string): HashlineToken[] {
-		if (this.#closed) throw new Error("HashlineTokenizer is closed; call reset() before reusing.");
-		if (chunk.length === 0) return [];
-		this.#buffer = this.#buffer ? this.#buffer + chunk : chunk;
-		return this.#drainCompleteLines();
-	}
-
-	/**
-	 * Flush any buffered residual line (the last line of input when it lacks
-	 * a trailing newline) and mark the tokenizer closed. Calling `end` a
-	 * second time returns `[]`; reuse requires `reset`.
-	 */
-	end(): HashlineToken[] {
-		if (this.#closed) return [];
-		this.#closed = true;
-		const buf = this.#buffer;
-		this.#buffer = "";
-		if (buf.length === 0) return [];
-		let stop = buf.length;
-		if (buf.charCodeAt(stop - 1) === CHAR_CARRIAGE_RETURN) stop--;
-		const token = classifyLine(buf.slice(0, stop), this.#nextLineNum++);
-		return [token];
-	}
-
-	/** Discard any buffered text and reset the line counter to 1. */
-	reset(): void {
-		this.#buffer = "";
-		this.#nextLineNum = 1;
-		this.#closed = false;
-	}
-
-	/** Convenience: feed an entire text and immediately flush. */
-	tokenizeAll(text: string): HashlineToken[] {
-		this.reset();
-		const first = this.feed(text);
-		const last = this.end();
-		return last.length === 0 ? first : first.concat(last);
-	}
-
-	/** Stateless one-shot classification. Does not touch the streaming buffer. */
-	tokenize(line: string, lineNum = 0): HashlineToken {
-		return classifyLine(line, lineNum);
-	}
-
-	isOp(line: string): boolean {
-		return tryParseOp(line) !== null;
-	}
-
-	isHeader(line: string): boolean {
-		return tryParseHeader(line) !== null;
-	}
-
-	isEnvelopeMarker(line: string): boolean {
-		return (
-			markerLineEquals(line, BEGIN_PATCH_MARKER) ||
-			markerLineEquals(line, END_PATCH_MARKER) ||
-			markerLineEquals(line, ABORT_MARKER)
-		);
-	}
-
-	#drainCompleteLines(): HashlineToken[] {
-		const tokens: HashlineToken[] = [];
-		const buf = this.#buffer;
-		let start = 0;
-		for (let index = 0; index < buf.length; index++) {
-			if (buf.charCodeAt(index) !== CHAR_LINE_FEED) continue;
-			let stop = index;
-			if (stop > start && buf.charCodeAt(stop - 1) === CHAR_CARRIAGE_RETURN) stop--;
-			tokens.push(classifyLine(buf.slice(start, stop), this.#nextLineNum++));
-			start = index + 1;
-		}
-		this.#buffer = start < buf.length ? buf.slice(start) : "";
-		return tokens;
-	}
-}

package/src/hashline/types.ts

@@ -1,66 +0,0 @@
-import * as z from "zod/v4";
-import type { LspBatchRequest } from "../edit/renderer";
-import type { WritethroughCallback, WritethroughDeferredHandle } from "../lsp";
-import type { ToolSession } from "../tools";
-
-export type Anchor = {
-	line: number;
-};
-
-export type HashlineCursor =
-	| { kind: "bof" }
-	| { kind: "eof" }
-	| { kind: "before_anchor"; anchor: Anchor }
-	| { kind: "after_anchor"; anchor: Anchor };
-
-export type HashlineEdit =
-	| { kind: "insert"; cursor: HashlineCursor; text: string; lineNum: number; index: number }
-	| { kind: "delete"; anchor: Anchor; lineNum: number; index: number; oldAssertion?: string };
-
-export interface HashlineInputSection {
-	path: string;
-	fileHash?: string;
-	diff: string;
-}
-
-/** `path` is accepted by the edit tool runtime; other extra keys are preserved. */
-export const hashlineEditParamsSchema = z.object({ input: z.string(), path: z.string().optional() }).passthrough();
-export type HashlineParams = z.infer<typeof hashlineEditParamsSchema>;
-
-export interface HashlineStreamOptions {
-	/** First line number to use when formatting (1-indexed). */
-	startLine?: number;
-	/** Maximum formatted lines per yielded chunk (default: 200). */
-	maxChunkLines?: number;
-	/** Maximum UTF-8 bytes per yielded chunk (default: 64 KiB). */
-	maxChunkBytes?: number;
-}
-
-export interface CompactHashlineDiffPreview {
-	preview: string;
-	addedLines: number;
-	removedLines: number;
-}
-
-export interface CompactHashlineDiffOptions {
-	/** Maximum entries kept on each side of an unchanged-context truncation (default: 2). */
-	maxUnchangedRun?: number;
-}
-export interface HashlineApplyOptions {
-	autoDropPureInsertDuplicates?: boolean;
-}
-
-export interface SplitHashlineOptions {
-	cwd?: string;
-	path?: string;
-}
-
-export interface ExecuteHashlineSingleOptions {
-	session: ToolSession;
-	input: string;
-	path?: string;
-	signal?: AbortSignal;
-	batchRequest?: LspBatchRequest;
-	writethrough: WritethroughCallback;
-	beginDeferredDiagnosticsForPath: (path: string) => WritethroughDeferredHandle;
-}

package/src/prompts/tools/hashline.md

@@ -1,83 +0,0 @@
-Your patch language is a compact, line-anchored edit format.
-
-<payload>
-Patch payload is a series of hunks: `¶PATH#HASH` header followed by any number of operations. `HASH` should be copied as is from read/search. Missing? Re-`read`.
-- No context rows, no gutters.
-- NEVER restate unchanged lines "for context".
-- Op lines carry NO payload. Every payload line lives on its own row and MUST start with `+`; that delimiter is stripped.
-- Payload indentation is literal.
-</payload>
-
-<ops>
-LINE↑    insert before (or BOF↑)
-LINE↓    insert after  (or EOF↓)
-A-B:     replace A..B  (or A: == A..A)
-A-B!     delete A..B   (or A! == A..A)
-+PAYLOAD payload line for the preceding op
-</ops>
-
-<rules>
-- **Payload is only what's NEW.** `:` replaces inside; `↑`/`↓` add at anchor. NEVER repeat anchor lines or neighbors.
-- **Use `+` for a blank payload line; use `++text` to write a line starting with `+text`.**
-- **Inserts add ONLY the rows you list.** The file's existing newlines around the anchor stay. NEVER tack a trailing `+` blank "for spacing" — it writes a literal blank line into the file, doubling whatever is already there.
-- **A bare `LINE↑`/`LINE↓` with no payload still inserts ONE blank line.** Not a no-op. Omit the op if you want nothing there.
-- **Go small.** Add → `↑`/`↓`; replace → `:`; delete → `!`.
-- **Line numbers are frozen references to what you have seen.** Later ops in the same hunk still use original line numbers; they do NOT shift as earlier ops apply.
-</rules>
-
-<common-failures>
-- **NEVER replay past your range.** Stop before B+1; extend B if needed.
-- **Read lines look like replace ops.** `84:content` = "make line 84 content" — and inline content is rejected. Don't echo read-style rows.
-- **NEVER fabricate file hashes.** Missing? Re-`read`.
-</common-failures>
-
-<example>
-```a.ts#1a2b
-1:const X = "a";
-2:
-3:export function f() { return X; }
-4:f();
-```
-
-# replace one line, insert after, delete
-```
-¶a.ts#1a2b
-1:
-+const X = "b";
-+export const Y = X;
-1↓
-+const Z = Y;
-4!
-```
-</example>
-
-<anti-pattern>
-# WRONG — inline payload after the sigil is rejected
-1:const X = "b";
-1↓const Z = Y;
-1-2:const X = "b";
-+export const Y = X;
-# WRONG — INSERT used to change a line (old line survives)
-1↓
-+const X = "b";
-# WRONG — echoing read-style lines as context before the real op
-1:const X = "a";
-1-2:
-+const X = "b";
-+export const Y = X;
-# WRONG — trailing `+` blank writes a literal empty line; the new blank lands right next to the orig blank at line 2, doubling it
-1↓
-+const Y = X;
-+
-# WRONG — `2↓` still anchors at PRE-EDIT line 2 (frozen), NOT at the line just inserted by `1↓`. Both inserts land at their own anchors, giving three consecutive blanks (new from `1↓`, orig blank line 2, new from `2↓`).
-1↓
-2↓
-</anti-pattern>
-
-<critical>
-- One op per range, ever.
-- Pick op precisely. Update: `:`, add: `↑`/`↓`, remove: `!`.
-- Payload always lives on its own `+`-prefixed line — never inline with the op.
-- Payload is only what's NEW; never repeat anchor lines or neighbors.
-- Anchor exactly; don't anchor neighbors.
-</critical>