dsh-hashline-edit 0.2.1 → 0.2.2

package/src/apply-patch/grammar.ts

@@ -1,620 +0,0 @@
-/**
- * apply_patch v4 — grammar layer.
- *
- * Owns the Codex patch syntax plus the two extensions:
- *
- * - fragment mode: `@@` / `@@ <context>` + ' '/'+'/'-' body lines (+ optional
- *   `*** End of File`),
- * - identity mode: `@@<a>` / `@@<a>-<b>` + '+' lines only (empty body = delete),
- * - reference hunks: `*** Copy From: <path> <a>-<b>` / `*** Cut From: ...`,
- * - file lifecycle: `*** Add File:`, `*** Delete File:`, `*** Move to:`.
- *
- * The layer is pure: it never touches the filesystem and never resolves anchors.
- * Every problem it reports carries the 1-based patch line, so the renderer can
- * print one rejection listing every syntax problem in the submission.
- *
- * @module dsh-hashline-edit/apply-patch/grammar
- */
-import { HASH_RE } from "../hashline/alphabet.js";
-import type {
-  AddFileOp,
-  AnchorSpan,
-  DeleteFileOp,
-  FragmentHunk,
-  FragmentLine,
-  HunkAddress,
-  IdentityHunk,
-  MoveDirective,
-  ParsePatchResult,
-  PatchDocument,
-  PatchFileOp,
-  PatchHunk,
-  PatchProblem,
-  ReferenceHunk,
-  ReferenceSource,
-  UpdateFileOp,
-} from "./types.js";
-
-/** Hard cap on one patch input (bytes), mirroring the host request budget. */
-export const PATCH_MAX_BYTES = 512 * 1024;
-
-/** Options for {@link parseApplyPatch}. */
-export interface ParsePatchOptions {
-  /** Maximum accepted input size in bytes; defaults to {@link PATCH_MAX_BYTES}. */
-  readonly maxBytes?: number;
-}
-
-const BEGIN_PATCH = "*** Begin Patch";
-const END_PATCH = "*** End Patch";
-const ADD_FILE = "*** Add File:";
-const DELETE_FILE = "*** Delete File:";
-const UPDATE_FILE = "*** Update File:";
-const MOVE_TO = "*** Move to:";
-const COPY_FROM = "*** Copy From:";
-const CUT_FROM = "*** Cut From:";
-const END_OF_FILE = "*** End of File";
-
-type ReferenceMode = "copy" | "cut";
-
-function syntaxProblem(patchLine: number, message: string, path?: string): PatchProblem {
-  if (path === undefined) return { code: "E_PATCH_SYNTAX", patchLine, message };
-  return { code: "E_PATCH_SYNTAX", patchLine, path, message };
-}
-
-function quote(line: string): string {
-  return JSON.stringify(line);
-}
-
-/**
- * Split a patch into physical lines while keeping the 1-based line numbers the
- * diagnostics report. A UTF-8 BOM and CRLF endings are accepted; the final
- * newline does not create a phantom line.
- */
-function splitPatchLines(input: string): string[] {
-  const text = input.charCodeAt(0) === 0xfeff ? input.slice(1) : input;
-  const lines = text.split("\n");
-  if (lines.length > 0 && lines[lines.length - 1] === "") lines.pop();
-  return lines.map((line) => (line.endsWith("\r") ? line.slice(0, -1) : line));
-}
-
-/**
- * Parse one patch document.
- *
- * Success returns the AST defined in `types.ts`; failure returns every syntax
- * problem the parser could diagnose in one pass (never a throw for malformed
- * user input). Rejections use `E_PATCH_SYNTAX` with the offending patch line.
- *
- * @param input - the raw `*** Begin Patch` … `*** End Patch` text.
- * @param options - size guard knobs.
- * @returns the parsed document or the collected problems.
- */
-export function parseApplyPatch(input: string, options?: ParsePatchOptions): ParsePatchResult {
-  const maxBytes = options?.maxBytes ?? PATCH_MAX_BYTES;
-  if (new TextEncoder().encode(input).byteLength > maxBytes) {
-    return {
-      ok: false,
-      problems: [syntaxProblem(1, `patch is larger than the ${maxBytes}-byte input limit`)],
-    };
-  }
-  return new PatchParser(splitPatchLines(input)).parse();
-}
-
-/**
- * Parse an anchor token: `<a>` (single line) or `<a>-<b>` (closed span).
- * Returns `undefined` when the token is not a valid four-character anchor pair; the
- * grammar layer turns that into an `E_PATCH_SYNTAX` problem at the patch line.
- *
- * @param token - the text after `@@` (no leading space).
- * @returns the closed span, or undefined for a malformed token.
- */
-export function parseAnchorSpan(token: string): AnchorSpan | undefined {
-  const parts = token.split("-");
-  if (parts.length === 1) {
-    const anchor = parts[0] ?? "";
-    if (!HASH_RE.test(anchor)) return undefined;
-    return { from: anchor, to: anchor };
-  }
-  if (parts.length === 2) {
-    const from = parts[0] ?? "";
-    const to = parts[1] ?? "";
-    if (!HASH_RE.test(from) || !HASH_RE.test(to)) return undefined;
-    return { from, to };
-  }
-  return undefined;
-}
-
-class PatchParser {
-  private index = 0;
-  private readonly problems: PatchProblem[] = [];
-  private readonly files: PatchFileOp[] = [];
-
-  constructor(private readonly lines: readonly string[]) {}
-
-  parse(): ParsePatchResult {
-    this.skipBlankLines();
-    if (this.atEnd()) {
-      return {
-        ok: false,
-        problems: [syntaxProblem(1, `patch is empty; expected '${BEGIN_PATCH}'`)],
-      };
-    }
-    const first = this.lines[this.index] ?? "";
-    if (first.trim() !== BEGIN_PATCH) {
-      return {
-        ok: false,
-        problems: [
-          syntaxProblem(
-            this.index + 1,
-            `the first non-empty line must be '${BEGIN_PATCH}', got ${quote(first)}`,
-          ),
-        ],
-      };
-    }
-    this.index += 1;
-
-    let sawEnd = false;
-    while (!this.atEnd()) {
-      const raw = this.lines[this.index] ?? "";
-      const trimmed = raw.trim();
-      if (trimmed === "") {
-        this.index += 1;
-        continue;
-      }
-      if (trimmed === END_PATCH) {
-        sawEnd = true;
-        this.index += 1;
-        break;
-      }
-      if (trimmed.startsWith("*** ")) {
-        const file = this.parseFileHeader(trimmed);
-        if (file !== undefined) this.files.push(file);
-        continue;
-      }
-      this.problem(
-        this.index,
-        `expected a file header ('*** Add File:', '*** Delete File:' or '*** Update File:'), got ${quote(raw)}`,
-      );
-      this.index += 1;
-    }
-
-    if (!sawEnd) {
-      this.problem(this.lines.length, `missing '${END_PATCH}'`);
-    } else {
-      while (!this.atEnd()) {
-        const raw = this.lines[this.index] ?? "";
-        if (raw.trim() !== "") {
-          this.problem(this.index, `unexpected content after '${END_PATCH}': ${quote(raw)}`);
-        }
-        this.index += 1;
-      }
-    }
-
-    if (this.problems.length > 0) return { ok: false, problems: this.problems };
-    const document: PatchDocument = { files: this.files, lineCount: this.lines.length };
-    return { ok: true, document };
-  }
-
-  private atEnd(): boolean {
-    return this.index >= this.lines.length;
-  }
-
-  private skipBlankLines(): void {
-    while (!this.atEnd() && (this.lines[this.index] ?? "").trim() === "") this.index += 1;
-  }
-
-  private problem(zeroBasedIndex: number, message: string): void {
-    this.problems.push(syntaxProblem(zeroBasedIndex + 1, message));
-  }
-
-  private parseFileHeader(trimmed: string): PatchFileOp | undefined {
-    const line = this.index + 1;
-    if (trimmed.startsWith(ADD_FILE)) return this.parseAddFile(line);
-    if (trimmed.startsWith(DELETE_FILE)) return this.parseDeleteFile(line);
-    if (trimmed.startsWith(UPDATE_FILE)) return this.parseUpdateFile(line);
-    if (trimmed === BEGIN_PATCH) {
-      this.problem(this.index, `duplicate '${BEGIN_PATCH}'`);
-      this.index += 1;
-      return undefined;
-    }
-    if (trimmed.startsWith(MOVE_TO) || trimmed === MOVE_TO) {
-      this.problem(
-        this.index,
-        `'${MOVE_TO}' is only valid directly after '${UPDATE_FILE}' and before the first hunk`,
-      );
-      this.index += 1;
-      return undefined;
-    }
-    if (trimmed.startsWith(COPY_FROM) || trimmed.startsWith(CUT_FROM)) {
-      this.problem(
-        this.index,
-        `'${COPY_FROM}' / '${CUT_FROM}' must be preceded by an identity hunk header ('@@<a>' or '@@<a>-<b>')`,
-      );
-      this.index += 1;
-      return undefined;
-    }
-    if (trimmed === END_OF_FILE) {
-      this.problem(this.index, `'${END_OF_FILE}' is only valid at the end of a fragment hunk`);
-      this.index += 1;
-      return undefined;
-    }
-    this.problem(this.index, `unknown patch directive ${quote(trimmed)}`);
-    this.index += 1;
-    return undefined;
-  }
-
-  private headerPath(trimmedLine: string, marker: string, patchLine: number): string | undefined {
-    const rest = trimmedLine.slice(marker.length);
-    if (rest === "") {
-      this.problem(patchLine - 1, `'${marker}' requires a non-empty path`);
-      return undefined;
-    }
-    if (!rest.startsWith(" ")) {
-      this.problem(patchLine - 1, `expected '${marker} <path>', got ${quote(trimmedLine)}`);
-      return undefined;
-    }
-    const path = rest.trim();
-    if (path === "") {
-      this.problem(patchLine - 1, `'${marker}' requires a non-empty path`);
-      return undefined;
-    }
-    return path;
-  }
-
-  private parseAddFile(headerLine: number): AddFileOp | undefined {
-    const path = this.headerPath(this.lines[this.index]?.trim() ?? "", ADD_FILE, headerLine);
-    this.index += 1;
-    const contentLines: string[] = [];
-    while (!this.atEnd()) {
-      const raw = this.lines[this.index] ?? "";
-      if (raw.trim().startsWith("*** ")) break;
-      if (raw.startsWith("+")) {
-        contentLines.push(raw.slice(1));
-        this.index += 1;
-        continue;
-      }
-      this.problem(this.index, `Add File content lines must start with '+', got ${quote(raw)}`);
-      this.index += 1;
-    }
-    if (path === undefined) return undefined;
-    return { kind: "add", path, headerLine, contentLines };
-  }
-
-  private parseDeleteFile(headerLine: number): DeleteFileOp | undefined {
-    const path = this.headerPath(this.lines[this.index]?.trim() ?? "", DELETE_FILE, headerLine);
-    this.index += 1;
-    while (!this.atEnd()) {
-      const raw = this.lines[this.index] ?? "";
-      const trimmed = raw.trim();
-      if (trimmed === "") {
-        this.index += 1;
-        continue;
-      }
-      if (trimmed.startsWith("*** ")) break;
-      this.problem(this.index, `Delete File takes no body lines, got ${quote(raw)}`);
-      this.index += 1;
-    }
-    if (path === undefined) return undefined;
-    return { kind: "delete", path, headerLine };
-  }
-
-  private parseUpdateFile(headerLine: number): UpdateFileOp | undefined {
-    const path = this.headerPath(this.lines[this.index]?.trim() ?? "", UPDATE_FILE, headerLine);
-    this.index += 1;
-    let moveTo: MoveDirective | undefined;
-    let firstHunkSeen = false;
-    const hunks: PatchHunk[] = [];
-
-    while (!this.atEnd()) {
-      const raw = this.lines[this.index] ?? "";
-      const trimmed = raw.trim();
-      if (trimmed === "") {
-        this.index += 1;
-        continue;
-      }
-      if (trimmed.startsWith("*** ")) {
-        if (trimmed.startsWith(MOVE_TO) || trimmed === MOVE_TO) {
-          const moveLine = this.index + 1;
-          const to = this.headerPath(trimmed, MOVE_TO, moveLine);
-          if (firstHunkSeen) {
-            this.problem(
-              this.index,
-              `'${MOVE_TO}' must appear directly after '${UPDATE_FILE}' and before the first hunk`,
-            );
-          } else if (moveTo !== undefined) {
-            this.problem(this.index, `duplicate '${MOVE_TO}' directive`);
-          } else if (to !== undefined) {
-            moveTo = { path: to, line: moveLine };
-          }
-          this.index += 1;
-          continue;
-        }
-        if (trimmed === END_OF_FILE) {
-          this.problem(this.index, `'${END_OF_FILE}' is only valid at the end of a fragment hunk`);
-          this.index += 1;
-          continue;
-        }
-        if (trimmed.startsWith(COPY_FROM) || trimmed.startsWith(CUT_FROM)) {
-          this.problem(
-            this.index,
-            `'${COPY_FROM}' / '${CUT_FROM}' must be preceded by an identity hunk header ('@@<a>' or '@@<a>-<b>')`,
-          );
-          this.index += 1;
-          continue;
-        }
-        break;
-      }
-      if (trimmed.startsWith("@@")) {
-        const hunk = this.parseHunk();
-        if (hunk !== undefined) hunks.push(hunk);
-        firstHunkSeen = true;
-        continue;
-      }
-      this.problem(
-        this.index,
-        `expected a hunk header '@@', '@@ <context>' or '@@<anchor>' before body lines, got ${quote(raw)}`,
-      );
-      this.index += 1;
-    }
-
-    if (path === undefined) return undefined;
-    const op: UpdateFileOp = {
-      kind: "update",
-      path,
-      headerLine,
-      hunks,
-      ...(moveTo === undefined ? {} : { moveTo }),
-    };
-    return op;
-  }
-
-  private parseHunk(): PatchHunk | undefined {
-    const headerLine = this.index + 1;
-    const header = this.lines[this.index] ?? "";
-    if (header === "@@" || header.startsWith("@@ ")) {
-      if (this.nextLineIsReference()) {
-        this.rejectFragmentReference(headerLine);
-        return undefined;
-      }
-      const address: HunkAddress =
-        header === "@@" ? { kind: "start" } : { kind: "context", text: header.slice(3) };
-      this.index += 1;
-      return this.parseFragmentBody(address, headerLine);
-    }
-
-    const token = header.slice(2).trimEnd();
-    const span = parseAnchorSpan(token);
-    if (span === undefined) {
-      this.problem(
-        this.index,
-        `invalid hunk header ${quote(header)}: expected '@@', '@@ <context>', '@@<anchor>' or '@@<anchor>-<anchor>'`,
-      );
-      this.index += 1;
-      this.skipLooseBodyLines();
-      return undefined;
-    }
-    const isReference = this.nextLineIsReference();
-    this.index += 1;
-    if (isReference) {
-      return this.parseReferenceBody({ kind: "anchors", span }, headerLine);
-    }
-    return this.parseIdentityBody(span, headerLine);
-  }
-
-  private nextLineIsReference(): boolean {
-    const next = this.lines[this.index + 1];
-    if (next === undefined) return false;
-    const trimmed = next.trim();
-    return (
-      trimmed === COPY_FROM ||
-      trimmed.startsWith(`${COPY_FROM} `) ||
-      trimmed === CUT_FROM ||
-      trimmed.startsWith(`${CUT_FROM} `)
-    );
-  }
-
-  private rejectFragmentReference(headerLine: number): void {
-    this.problem(
-      headerLine - 1,
-      `reference hunks require identity addressing ('@@<a>' or '@@<a>-<b>'); fragment addressing is not allowed before '${COPY_FROM}' / '${CUT_FROM}'`,
-    );
-    this.index += 2;
-    this.skipLooseBodyLines();
-  }
-
-  private skipLooseBodyLines(): void {
-    while (!this.atEnd()) {
-      const raw = this.lines[this.index] ?? "";
-      if (raw.trim() === "") {
-        this.index += 1;
-        continue;
-      }
-      if (raw.startsWith("@@") || raw.startsWith("*** ")) return;
-      this.index += 1;
-    }
-  }
-
-  private parseFragmentBody(address: HunkAddress, headerLine: number): FragmentHunk {
-    const lines: FragmentLine[] = [];
-    let endOfFile: number | undefined;
-    let ended = false;
-    while (!this.atEnd()) {
-      const raw = this.lines[this.index] ?? "";
-      if (ended) {
-        const trimmed = raw.trim();
-        if (trimmed === "") {
-          this.index += 1;
-          continue;
-        }
-        if (raw.startsWith("@@") || raw.startsWith("*** ")) break;
-        this.problem(this.index, `unexpected line after '${END_OF_FILE}': ${quote(raw)}`);
-        this.index += 1;
-        continue;
-      }
-      if (raw === "") {
-        this.problem(
-          this.index,
-          `fragment hunk body lines must start with ' ', '-' or '+', got ${quote(raw)}`,
-        );
-        this.index += 1;
-        continue;
-      }
-      if (raw.startsWith(" ")) {
-        lines.push({ op: "context", text: raw.slice(1), line: this.index + 1 });
-        this.index += 1;
-        continue;
-      }
-      if (raw.startsWith("-")) {
-        lines.push({ op: "remove", text: raw.slice(1), line: this.index + 1 });
-        this.index += 1;
-        continue;
-      }
-      if (raw.startsWith("+")) {
-        lines.push({ op: "add", text: raw.slice(1), line: this.index + 1 });
-        this.index += 1;
-        continue;
-      }
-      if (raw.trimEnd() === END_OF_FILE) {
-        endOfFile = this.index + 1;
-        this.problemIfEmptyFragment(lines, endOfFile);
-        ended = true;
-        this.index += 1;
-        continue;
-      }
-      if (raw.startsWith("@@") || raw.startsWith("*** ")) break;
-      this.problem(
-        this.index,
-        `invalid fragment body line ${quote(raw)}: every body line must start with ' ', '-' or '+'`,
-      );
-      this.index += 1;
-    }
-    if (lines.length === 0 && endOfFile === undefined) {
-      this.problem(headerLine - 1, "fragment hunk requires at least one body line");
-    }
-    return {
-      kind: "fragment",
-      address,
-      headerLine,
-      lines,
-      ...(endOfFile === undefined ? {} : { endOfFile }),
-    };
-  }
-
-  private problemIfEmptyFragment(lines: readonly FragmentLine[], markerLine: number): void {
-    if (lines.length === 0) {
-      this.problem(markerLine - 1, "fragment hunk requires at least one body line");
-    }
-  }
-
-  private parseIdentityBody(span: AnchorSpan, headerLine: number): IdentityHunk {
-    const addedLines: string[] = [];
-    while (!this.atEnd()) {
-      const raw = this.lines[this.index] ?? "";
-      if (raw === "") break;
-      if (raw.startsWith(" ")) {
-        this.problem(
-          this.index,
-          `identity hunk only allows '+' body lines; found a context line ${quote(raw)}`,
-        );
-        this.index += 1;
-        continue;
-      }
-      if (raw.startsWith("-")) {
-        this.problem(
-          this.index,
-          `identity hunk only allows '+' body lines; found a '-' line ${quote(raw)} (an empty body deletes the span)`,
-        );
-        this.index += 1;
-        continue;
-      }
-      if (raw.startsWith("+")) {
-        addedLines.push(raw.slice(1));
-        this.index += 1;
-        continue;
-      }
-      if (raw.trimEnd() === END_OF_FILE) {
-        this.problem(this.index, `'${END_OF_FILE}' is only valid in a fragment hunk`);
-        this.index += 1;
-        continue;
-      }
-      if (raw.startsWith("@@") || raw.startsWith("*** ")) break;
-      this.problem(
-        this.index,
-        `invalid identity body line ${quote(raw)}: identity hunks allow only '+' lines`,
-      );
-      this.index += 1;
-    }
-    return { kind: "identity", span, headerLine, addedLines };
-  }
-
-  private parseReferenceBody(address: HunkAddress, headerLine: number): ReferenceHunk | undefined {
-    if (address.kind !== "anchors") {
-      this.problem(headerLine - 1, "reference hunks require identity addressing");
-      return undefined;
-    }
-    const sourceLine = this.lines[this.index] ?? "";
-    const sourceLineNumber = this.index + 1;
-    const trimmed = sourceLine.trim();
-    const mode: ReferenceMode = trimmed.startsWith(CUT_FROM) ? "cut" : "copy";
-    const source = this.parseReferenceSource(trimmed, sourceLineNumber, mode);
-    this.index += 1;
-    while (!this.atEnd()) {
-      const raw = this.lines[this.index] ?? "";
-      if (raw.trim() === "") {
-        this.index += 1;
-        continue;
-      }
-      if (raw.startsWith("@@") || raw.startsWith("*** ")) break;
-      this.problem(this.index, `reference hunk takes no body lines, got ${quote(raw)}`);
-      this.index += 1;
-    }
-    if (source === undefined) return undefined;
-    return { kind: "reference", address, source, headerLine };
-  }
-
-  private parseReferenceSource(
-    trimmedLine: string,
-    patchLine: number,
-    mode: ReferenceMode,
-  ): ReferenceSource | undefined {
-    const marker = mode === "copy" ? COPY_FROM : CUT_FROM;
-    const rest = trimmedLine.slice(marker.length);
-    if (rest === "") {
-      this.problem(
-        patchLine - 1,
-        `'${marker}' requires a source path and an '<anchor>-<anchor>' span`,
-      );
-      return undefined;
-    }
-    if (!rest.startsWith(" ")) {
-      this.problem(
-        patchLine - 1,
-        `expected '${marker} <path> <anchor>-<anchor>', got ${quote(trimmedLine)}`,
-      );
-      return undefined;
-    }
-    const body = rest.trim();
-    const split = body.lastIndexOf(" ");
-    if (split <= 0) {
-      this.problem(
-        patchLine - 1,
-        `expected '<path> <anchor>-<anchor>' after '${marker}' in ${quote(trimmedLine)}`,
-      );
-      return undefined;
-    }
-    const path = body.slice(0, split).trim();
-    const token = body.slice(split + 1);
-    if (path === "") {
-      this.problem(patchLine - 1, `'${marker}' requires a non-empty source path`);
-      return undefined;
-    }
-    const span = parseAnchorSpan(token);
-    if (span === undefined || !token.includes("-")) {
-      this.problem(
-        patchLine - 1,
-        `invalid reference span ${quote(token)}: expected '<anchor>-<anchor>'`,
-      );
-      return undefined;
-    }
-    return { mode, path, span, line: patchLine };
-  }
-}

package/src/apply-patch/index.ts

@@ -1,17 +0,0 @@
-/**
- * apply_patch v4 — public module surface.
- *
- * This barrel is the stable import point for the apply-patch lanes and the dsh
- * adapter. The individual modules keep their own boundaries; re-exporting them
- * here lets downstream code depend on one specifier and keeps the knip entry
- * graph honest while the feature is still being wired.
- *
- * @module dsh-hashline-edit/apply-patch
- */
-export * from "./types.js";
-export * from "./grammar.js";
-export * from "./resolve.js";
-export * from "./plan.js";
-export * from "./commit.js";
-export * from "./render.js";
-export * from "./tool.js";