@speclynx/apidom-parser-adapter-yaml-1-2 6.0.6 → 6.1.0

package/src/tree-sitter/syntactic-analysis/ast/schemas/canonical-format.mjs

@@ -1,15 +1,16 @@
-import { tail, compose, pathOr, map, concat, transduce, pipe, trim, split, join, curry } from 'ramda';
-import { isInteger, trimStart, trimEnd, isUndefined, trimCharsStart, isEmptyString, repeatStr, concatRight } from 'ramda-adjunct';
-import { unraw } from 'unraw';
+import { tail, pathOr, pipe, trim, curry, replace } from 'ramda';
+import { isInteger, isEmptyString, repeatStr } from 'ramda-adjunct';
 
 /**
  * Helpers.
  */
 
-const blockStyleRegExp = /^(?<style>[|>])(?<chomping>[+-]?)(?<indentation>[0-9]*)\s/;
+// indentation and chomping indicators may appear in either order (YAML 1.2 §8.1.1)
+const blockStyleRegExp = /^(?<style>[|>])(?<indicators>[+-]?[0-9]?|[0-9][+-])(?:\s|$)/;
 const getIndentationIndicator = content => {
   const matches = content.match(blockStyleRegExp);
-  const indicator = pathOr('', ['groups', 'indentation'], matches);
+  const indicators = pathOr('', ['groups', 'indicators'], matches);
+  const indicator = indicators.replace(/[+-]/, '');
   return isEmptyString(indicator) ? undefined : parseInt(indicator, 10);
 };
 const getIndentation = content => {
@@ -20,30 +21,72 @@ const getIndentation = content => {
     return repeatStr(' ', explicitIndentationIndicator);
   }
 
-  // we assume indentation indicator from first line
-  const firstLine = pathOr('', [1], content.split('\n'));
+  // we assume indentation indicator from first non-empty line (YAML 1.2 §8.1.1.1);
+  // with no non-empty line, the first line is used so that its spaces are still stripped
+  const lines = tail(content.split('\n'));
+  const firstLine = lines.find(line => !/^ *\r?$/.test(line)) ?? pathOr('', [0], lines);
   const implicitIndentationIndicator = pathOr(0, ['groups', 'indentation', 'length'], firstLine.match(/^(?<indentation>[ ]*)/));
   return repeatStr(' ', implicitIndentationIndicator);
 };
+
+// splits a block scalar body into deindented content lines and trailing lines (YAML 1.2 §8.1.1.2):
+// trailing lines are the whitespace-only lines after the last content line, minus any up to and
+// including the last one indented deeper than the block, which are content; without any content
+// line, all lines are trailing
+const splitBlockScalar = (content, chomping) => {
+  const indentation = getIndentation(content).length;
+  const normalized = normalizeLineBreaks(content);
+  const body = normalized.includes('\n') ? normalized.slice(normalized.indexOf('\n') + 1) : '';
+  const lines = body === '' ? [] : body.split('\n');
+  const deindent = line => line.slice(Math.min(indentation, line.search(/[^ ]|$/)));
+
+  // without keep chomping, trailing whitespace-only lines not indented deeper than the first
+  // content line don't belong to the block scalar
+  const firstContent = lines.find(line => /[^ ]/.test(line));
+  if (chomping !== '+' && firstContent !== undefined) {
+    const contentIndentation = firstContent.search(/[^ ]/);
+    while (/^ *$/.test(lines.at(-1)) && lines.at(-1).length <= contentIndentation) lines.pop();
+  }
+  let chompStart = lines.length;
+  while (chompStart > 0 && /^ *$/.test(lines[chompStart - 1])) chompStart -= 1;
+  if (chompStart === 0) return {
+    lines: [],
+    trailing: lines.map(deindent)
+  };
+  for (let i = lines.length - 1; i >= chompStart; i -= 1) {
+    if (lines[i].length > indentation) {
+      chompStart = i + 1;
+      break;
+    }
+  }
+  return {
+    lines: lines.slice(0, chompStart).map(deindent),
+    trailing: lines.slice(chompStart).map(deindent)
+  };
+};
 const getChompingIndicator = content => {
   const matches = content.match(blockStyleRegExp);
-  const indicator = pathOr('', ['groups', 'chomping'], matches);
+  const indicators = pathOr('', ['groups', 'indicators'], matches);
+  const indicator = indicators.replace(/[0-9]/, '');
   return isEmptyString(indicator) ? undefined : indicator;
 };
-const chomp = (indicator, content) => {
-  // clip (single newline at end)
-  if (isUndefined(indicator)) {
-    return `${trimEnd(content)}\n`;
+const chomp = (indicator, value, {
+  lines,
+  trailing
+}) => {
+  // a block without content is empty, unless keep chomping preserves its line breaks
+  if (lines.length === 0) {
+    return indicator === '+' && trailing.length > 0 ? '\n'.repeat(Math.max(1, trailing.length - 1)) : '';
   }
   // strip (no newline at end)
-  if (indicator === '-') {
-    return trimEnd(content);
-  }
-  // keep (all newlines from end)
+  if (indicator === '-') return value;
+  // keep (trailing lines and a newline at end)
   if (indicator === '+') {
-    return content;
+    const kept = value + trailing.map(line => `\n${line}`).join('');
+    return kept.endsWith('\n') ? kept : `${kept}\n`;
   }
-  return content;
+  // clip (single newline at end)
+  return `${value}\n`;
 };
 
 /**
@@ -53,56 +96,108 @@ const chomp = (indicator, content) => {
 // @ts-ignore
 const normalizeLineBreaks = val => val.replace(/\r\n/g, '\n');
 
-// prevent escaped line breaks from being converted to a space
-const preventLineBreakCollapseToSpace = val => val.replace(/\\\n\s*/g, '');
-
-// collapse line breaks into spaces
-const collapseLineBreakToSpace = val => {
-  /**
-   * Safari doesn't support negative lookbehind, thus we use mimicking technique:
-   *
-   * - https://blog.stevenlevithan.com/archives/mimic-lookbehind-javascript
-   *
-   * Ideally we want to use following replace, but that's not currently possible:
-   *
-   * .replace(/[^\n]\n([^\n]+)/g, (match: string, p1: string) => ` ${p1.trimLeft()}`)
-   */
-  return val.replace(/(\n)?\n([^\n]+)/g, (match, p1, p2) => p1 ? match : ` ${p2.trimStart()}`).replace(/[\n]{2}/g, '\n');
+// folds the line breaks of a flow scalar (YAML 1.2 §6.5, §7.3):
+// white space around a line break is dropped, a single line break becomes a space and every
+// further one a line break; with escapes (double-quoted style), escape sequences are left for
+// unescaping and an escaped line break is dropped together with the next line's indentation
+const foldFlowLines = curry((escapes, text) => text.replace(
+// a white space run not followed by a line break is consumed whole, which keeps the search
+// linear instead of rescanning the run from every position in it
+escapes ? /\\(?:\n[ \t]*|[^])|[ \t]*\n(?:[ \t]*\n)*[ \t]*|[ \t]+/g : /[ \t]*\n(?:[ \t]*\n)*[ \t]*|[ \t]+/g, match => {
+  if (match.startsWith('\\')) return match[1] === '\n' ? '' : match;
+  if (!match.includes('\n')) return match;
+  const lineBreaks = match.split('\n').length - 1;
+  return lineBreaks === 1 ? ' ' : '\n'.repeat(lineBreaks - 1);
+}));
+
+// folds the line break between two lines into a space; empty lines between them stay as line
+// breaks, and a line break next to a more-indented line isn't folded (YAML 1.2 §6.5)
+const foldLines = text => {
+  const parts = text.split(/\n(\n*)/); // lines alternating with the empty lines between them
+  let folded = parts[0];
+  for (let i = 1; i < parts.length; i += 2) {
+    const moreIndented = /^[ \t]/.test(parts[i - 1]) || /^[ \t]/.test(parts[i + 1]);
+    folded += (moreIndented ? `\n${parts[i]}` : parts[i] || ' ') + parts[i + 1];
+  }
+  return folded;
+};
+
+// escaped characters of YAML 1.2 §5.7, productions [42] to [58]
+const escapedCharacters = {
+  '0': '\u0000',
+  // null
+  a: '\u0007',
+  // bell
+  b: '\u0008',
+  // backspace
+  t: '\u0009',
+  // horizontal tab
+  '\t': '\u0009',
+  // horizontal tab
+  n: '\u000a',
+  // line feed
+  v: '\u000b',
+  // vertical tab
+  f: '\u000c',
+  // form feed
+  r: '\u000d',
+  // carriage return
+  e: '\u001b',
+  // escape
+  ' ': '\u0020',
+  // space
+  '"': '\u0022',
+  // double quote
+  '/': '\u002f',
+  // slash
+  '\\': '\u005c',
+  // backslash
+  N: '\u0085',
+  // next line
+  _: '\u00a0',
+  // non-breaking space
+  L: '\u2028',
+  // line separator
+  P: '\u2029' // paragraph separator
 };
+
+// unescapes the escape sequences of a double-quoted scalar (YAML 1.2 §5.7); an escape sequence
+// outside of the spec is kept as-is
+const unescapeDoubleQuoted = text => text.replace(/\\(?:x([0-9a-fA-F]{2})|u([0-9a-fA-F]{4})|U([0-9a-fA-F]{8})|([^]))/g, (match, x, u, U, char) => {
+  const hex = x ?? u ?? U;
+  if (hex === undefined) return escapedCharacters[char] ?? match;
+  const codePoint = parseInt(hex, 16);
+  return codePoint <= 0x10ffff ? String.fromCodePoint(codePoint) : match;
+});
 const removeQuotes = curry((quoteType, val) => val.replace(new RegExp(`^${quoteType}`), '').replace(new RegExp(`${quoteType}$`), ''));
 
 /**
  * Formats Flow Scalar Plain style.
  * https://yaml.org/spec/1.2/spec.html#id2788859
  */
-export const formatFlowPlain = pipe(normalizeLineBreaks, trim, collapseLineBreakToSpace, split('\n'), map(trimStart), join('\n'));
+export const formatFlowPlain = pipe(normalizeLineBreaks, trim, foldFlowLines(false));
 
 /**
  * Formats Flow Scalar Single-Quoted style.
  * https://yaml.org/spec/1.2/spec.html#id2788097
  */
 
-export const formatFlowSingleQuoted = pipe(normalizeLineBreaks, trim, collapseLineBreakToSpace, split('\n'), map(trimStart), join('\n'), removeQuotes("'"));
+export const formatFlowSingleQuoted = pipe(normalizeLineBreaks, trim, foldFlowLines(false), removeQuotes("'"), replace(/''/g, "'"));
 
 /**
  * Formats Flow Scalar Double-Quoted style.
  * https://yaml.org/spec/1.2/spec.html#id2787109
  */
-export const formatFlowDoubleQuoted = pipe(normalizeLineBreaks, trim, preventLineBreakCollapseToSpace, collapseLineBreakToSpace, unraw, split('\n'), map(trimStart), join('\n'), removeQuotes('"'));
+export const formatFlowDoubleQuoted = pipe(normalizeLineBreaks, trim, foldFlowLines(true), unescapeDoubleQuoted, removeQuotes('"'));
 
 /**
  * Formats Block Scalar Literal style.
  * https://yaml.org/spec/1.2/spec.html#id2795688
  */
 export const formatBlockLiteral = content => {
-  const indentation = getIndentation(content);
-  const chompingIndicator = getChompingIndicator(content);
-  const normalized = normalizeLineBreaks(content);
-  const lines = tail(normalized.split('\n')); // first line only contains indicators
-  const transducer = compose(map(trimCharsStart(indentation)), map(concatRight('\n')));
-  // @ts-ignore
-  const deindented = transduce(transducer, concat, '', lines);
-  return chomp(chompingIndicator, deindented);
+  const chomping = getChompingIndicator(content);
+  const split = splitBlockScalar(content, chomping);
+  return chomp(chomping, split.lines.join('\n'), split);
 };
 
 /**
@@ -110,13 +205,11 @@ export const formatBlockLiteral = content => {
  * https://yaml.org/spec/1.2/spec.html#id2796251
  */
 export const formatBlockFolded = content => {
-  const indentation = getIndentation(content);
-  const chompingIndicator = getChompingIndicator(content);
-  const normalized = normalizeLineBreaks(content);
-  const lines = tail(normalized.split('\n')); // first line only contains indicators
-  const transducer = compose(map(trimCharsStart(indentation)), map(concatRight('\n')));
-  // @ts-ignore
-  const deindented = transduce(transducer, concat, '', lines);
-  const collapsed = collapseLineBreakToSpace(deindented);
-  return chomp(chompingIndicator, collapsed);
+  const chomping = getChompingIndicator(content);
+  const split = splitBlockScalar(content, chomping);
+  // leading line breaks are preserved as-is; folding applies only between content lines
+  const text = split.lines.join('\n');
+  const leadingLineBreaks = text.match(/^\n*/)[0];
+  const folded = leadingLineBreaks + foldLines(text.slice(leadingLineBreaks.length));
+  return chomp(chomping, folded, split);
 };