neon 3.0.0 → 3.1.1

package/dist/psql/complete/matcher.js

@@ -1,315 +1,251 @@
+//#region src/psql/complete/matcher.ts
 /**
- * Matches / TailMatches / HeadMatches DSL.
- *
- * TypeScript port of the pattern-matching helpers in psql's tab-complete.in.c.
- * The C source uses a varargs macro layered on top of `word_matches()` to
- * decide whether a sequence of "previous words" matches a pattern. We
- * reproduce the same shape with plain functions because (a) JS has no
- * preprocessor macros and (b) array spread is the natural varargs analogue.
- *
- *   Matches('SELECT', MatchAny, 'FROM')       — exact word-count match.
- *   TailMatches('SELECT', MatchAny, 'FROM')   — match the last N words.
- *   HeadMatches('ALTER', 'TABLE')             — match the first N words.
- *
- * Wildcards and special tokens (mirroring the upstream constants):
- *
- *   MatchAny             — any single word (`NULL` upstream).
- *   MatchAnyExcept(x)    — any word that doesn't match `x` (upstream `"!x"`).
- *   '*' inside a pattern — wildcard within a single word (e.g. `'pg_*'`).
- *   '|' inside a pattern — alternation (e.g. `'TABLE|VIEW'`).
- *
- * Comparisons are case-insensitive by default (SQL keywords); pass
- * `caseSensitive: true` to the DSL helpers when matching strictly.
- *
- * The tokenizer is exported separately because the `index.ts` entry point
- * needs to do the same word-split-then-match dance.
- */
-export const MatchAny = null;
-export const MatchAnyExcept = (pattern) => "!" + pattern;
+* Matches / TailMatches / HeadMatches DSL.
+*
+* TypeScript port of the pattern-matching helpers in psql's tab-complete.in.c.
+* The C source uses a varargs macro layered on top of `word_matches()` to
+* decide whether a sequence of "previous words" matches a pattern. We
+* reproduce the same shape with plain functions because (a) JS has no
+* preprocessor macros and (b) array spread is the natural varargs analogue.
+*
+*   Matches('SELECT', MatchAny, 'FROM')       — exact word-count match.
+*   TailMatches('SELECT', MatchAny, 'FROM')   — match the last N words.
+*   HeadMatches('ALTER', 'TABLE')             — match the first N words.
+*
+* Wildcards and special tokens (mirroring the upstream constants):
+*
+*   MatchAny             — any single word (`NULL` upstream).
+*   MatchAnyExcept(x)    — any word that doesn't match `x` (upstream `"!x"`).
+*   '*' inside a pattern — wildcard within a single word (e.g. `'pg_*'`).
+*   '|' inside a pattern — alternation (e.g. `'TABLE|VIEW'`).
+*
+* Comparisons are case-insensitive by default (SQL keywords); pass
+* `caseSensitive: true` to the DSL helpers when matching strictly.
+*
+* The tokenizer is exported separately because the `index.ts` entry point
+* needs to do the same word-split-then-match dance.
+*/
+const MatchAny = null;
+const MatchAnyExcept = (pattern) => "!" + pattern;
 const cimatch = (s1, s2, n, caseSensitive) => {
-    if (s1.length < n || s2.length < n)
-        return false;
-    const a = s1.slice(0, n);
-    const b = s2.slice(0, n);
-    if (caseSensitive)
-        return a === b;
-    return a.toLowerCase() === b.toLowerCase();
+	if (s1.length < n || s2.length < n) return false;
+	const a = s1.slice(0, n);
+	const b = s2.slice(0, n);
+	if (caseSensitive) return a === b;
+	return a.toLowerCase() === b.toLowerCase();
 };
 /**
- * Return true iff `word` matches `pattern`. `null`/`MatchAny` matches
- * everything. A leading `!` inverts the match. `|` separates alternatives.
- * `*` is a single-word wildcard.
- */
-export const wordMatches = (pattern, word, caseSensitive = false) => {
-    if (pattern === null)
-        return true;
-    if (pattern.startsWith("!")) {
-        return !wordMatches(pattern.slice(1), word, caseSensitive);
-    }
-    const wordlen = word.length;
-    let cursor = pattern;
-    for (;;) {
-        let starIdx = -1;
-        let i = 0;
-        while (i < cursor.length && cursor[i] !== "|") {
-            if (cursor[i] === "*")
-                starIdx = i;
-            i++;
-        }
-        if (starIdx >= 0) {
-            const beforeLen = starIdx;
-            const afterLen = i - starIdx - 1;
-            if (wordlen >= beforeLen + afterLen &&
-                cimatch(word, cursor, beforeLen, caseSensitive) &&
-                cimatch(word.slice(wordlen - afterLen), cursor.slice(starIdx + 1), afterLen, caseSensitive)) {
-                return true;
-            }
-        }
-        else {
-            if (wordlen === i &&
-                cimatch(word, cursor, wordlen, caseSensitive)) {
-                return true;
-            }
-        }
-        if (i >= cursor.length)
-            break;
-        cursor = cursor.slice(i + 1);
-    }
-    return false;
+* Return true iff `word` matches `pattern`. `null`/`MatchAny` matches
+* everything. A leading `!` inverts the match. `|` separates alternatives.
+* `*` is a single-word wildcard.
+*/
+const wordMatches = (pattern, word, caseSensitive = false) => {
+	if (pattern === null) return true;
+	if (pattern.startsWith("!")) return !wordMatches(pattern.slice(1), word, caseSensitive);
+	const wordlen = word.length;
+	let cursor = pattern;
+	for (;;) {
+		let starIdx = -1;
+		let i = 0;
+		while (i < cursor.length && cursor[i] !== "|") {
+			if (cursor[i] === "*") starIdx = i;
+			i++;
+		}
+		if (starIdx >= 0) {
+			const beforeLen = starIdx;
+			const afterLen = i - starIdx - 1;
+			if (wordlen >= beforeLen + afterLen && cimatch(word, cursor, beforeLen, caseSensitive) && cimatch(word.slice(wordlen - afterLen), cursor.slice(starIdx + 1), afterLen, caseSensitive)) return true;
+		} else if (wordlen === i && cimatch(word, cursor, wordlen, caseSensitive)) return true;
+		if (i >= cursor.length) break;
+		cursor = cursor.slice(i + 1);
+	}
+	return false;
 };
 /**
- * Exact-length match: every word in `words` must match the corresponding
- * pattern. `words` here is "previous words" — i.e. the words BEFORE the
- * current (in-progress) word.
- *
- * The upstream uses `previous_words_count + 1 == narg` because the C array
- * includes both prev words AND the current. Our representation only stores
- * prev words; the caller passes the in-progress word separately.
- *
- *   prev = ['SELECT'], pattern = ['SELECT']         → true   (1 == 1)
- *   prev = ['SELECT', 'a'], pattern = ['SELECT']    → false  (2 != 1)
- */
-export const Matches = (prev, patterns, caseSensitive = false) => {
-    if (prev.length !== patterns.length)
-        return false;
-    for (let k = 0; k < patterns.length; k++) {
-        if (!wordMatches(patterns[k], prev[k], caseSensitive))
-            return false;
-    }
-    return true;
+* Exact-length match: every word in `words` must match the corresponding
+* pattern. `words` here is "previous words" — i.e. the words BEFORE the
+* current (in-progress) word.
+*
+* The upstream uses `previous_words_count + 1 == narg` because the C array
+* includes both prev words AND the current. Our representation only stores
+* prev words; the caller passes the in-progress word separately.
+*
+*   prev = ['SELECT'], pattern = ['SELECT']         → true   (1 == 1)
+*   prev = ['SELECT', 'a'], pattern = ['SELECT']    → false  (2 != 1)
+*/
+const Matches = (prev, patterns, caseSensitive = false) => {
+	if (prev.length !== patterns.length) return false;
+	for (let k = 0; k < patterns.length; k++) if (!wordMatches(patterns[k], prev[k], caseSensitive)) return false;
+	return true;
 };
 /**
- * Tail match: the LAST N words of `prev` must match the patterns.
- *
- *   prev = ['SELECT', 'a', 'FROM'], patterns = ['FROM']   → true
- *   prev = ['SELECT'], patterns = ['SELECT', 'FROM']     → false (too few)
- */
-export const TailMatches = (prev, patterns, caseSensitive = false) => {
-    if (prev.length < patterns.length)
-        return false;
-    const offset = prev.length - patterns.length;
-    for (let k = 0; k < patterns.length; k++) {
-        if (!wordMatches(patterns[k], prev[offset + k], caseSensitive))
-            return false;
-    }
-    return true;
+* Tail match: the LAST N words of `prev` must match the patterns.
+*
+*   prev = ['SELECT', 'a', 'FROM'], patterns = ['FROM']   → true
+*   prev = ['SELECT'], patterns = ['SELECT', 'FROM']     → false (too few)
+*/
+const TailMatches = (prev, patterns, caseSensitive = false) => {
+	if (prev.length < patterns.length) return false;
+	const offset = prev.length - patterns.length;
+	for (let k = 0; k < patterns.length; k++) if (!wordMatches(patterns[k], prev[offset + k], caseSensitive)) return false;
+	return true;
 };
 /**
- * Head match: the FIRST N words of `prev` must match.
- *
- *   prev = ['ALTER', 'TABLE', 'foo'], patterns = ['ALTER', 'TABLE']  → true
- */
-export const HeadMatches = (prev, patterns, caseSensitive = false) => {
-    if (prev.length < patterns.length)
-        return false;
-    for (let k = 0; k < patterns.length; k++) {
-        if (!wordMatches(patterns[k], prev[k], caseSensitive))
-            return false;
-    }
-    return true;
+* Head match: the FIRST N words of `prev` must match.
+*
+*   prev = ['ALTER', 'TABLE', 'foo'], patterns = ['ALTER', 'TABLE']  → true
+*/
+const HeadMatches = (prev, patterns, caseSensitive = false) => {
+	if (prev.length < patterns.length) return false;
+	for (let k = 0; k < patterns.length; k++) if (!wordMatches(patterns[k], prev[k], caseSensitive)) return false;
+	return true;
 };
 /**
- * Split `input` into psql-style words for completion purposes.
- *
- *   - Whitespace is the basic separator.
- *   - "..."           treated as ONE word (quoted identifier).
- *   - '...'           treated as ONE word (string literal).
- *   - \backslash...   treated as ONE word (the backslash command).
- *   - Punctuation (commas, parens, semicolons) is broken into its own word.
- *   - Embedded `.` (schema-qualifier) keeps the dotted name as a single word
- *     so completion sees `pg_catalog.pg_class` whole.
- *
- * The tokenizer is intentionally pragmatic — it doesn't validate SQL, it
- * just slices well enough for the rule body to inspect previous words.
- */
-export const tokenize = (input) => {
-    const out = [];
-    let i = 0;
-    while (i < input.length) {
-        const ch = input[i];
-        // Whitespace.
-        if (ch === " " || ch === "\t" || ch === "\n" || ch === "\r") {
-            i++;
-            continue;
-        }
-        // Backslash command. Capture the entire backslash word as one token.
-        if (ch === "\\") {
-            const start = i;
-            i++;
-            // Single-char commands like `\!` and `\?` are valid; otherwise read
-            // letters until we hit whitespace or punctuation.
-            if (i < input.length && /[!?]/.test(input[i])) {
-                i++;
-            }
-            else {
-                while (i < input.length && /[A-Za-z_]/.test(input[i]))
-                    i++;
-                // Fold a trailing `+` run (the verbose suffix) into the command word
-                // so `\dt+ <TAB>` tokenizes as a single word and the describe rules
-                // that key on `prevWords.length === 1` still fire. The
-                // `S` (system-objects) suffix is already a letter, so it's consumed
-                // above; only `+` needs explicit handling here.
-                while (i < input.length && input[i] === "+")
-                    i++;
-            }
-            out.push({
-                text: input.slice(start, i),
-                raw: input.slice(start, i),
-                start,
-                end: i,
-            });
-            continue;
-        }
-        // Double-quoted identifier.
-        if (ch === '"') {
-            const start = i;
-            i++;
-            while (i < input.length) {
-                if (input[i] === '"') {
-                    // "" inside quotes is an escaped quote.
-                    if (input[i + 1] === '"') {
-                        i += 2;
-                        continue;
-                    }
-                    i++;
-                    break;
-                }
-                i++;
-            }
-            out.push({
-                text: input.slice(start, i),
-                raw: input.slice(start, i),
-                start,
-                end: i,
-            });
-            continue;
-        }
-        // Single-quoted string literal.
-        if (ch === "'") {
-            const start = i;
-            i++;
-            while (i < input.length) {
-                if (input[i] === "\\" && i + 1 < input.length) {
-                    i += 2;
-                    continue;
-                }
-                if (input[i] === "'") {
-                    if (input[i + 1] === "'") {
-                        i += 2;
-                        continue;
-                    }
-                    i++;
-                    break;
-                }
-                i++;
-            }
-            out.push({
-                text: input.slice(start, i),
-                raw: input.slice(start, i),
-                start,
-                end: i,
-            });
-            continue;
-        }
-        // Punctuation that splits words.
-        if (ch === "," ||
-            ch === ";" ||
-            ch === "(" ||
-            ch === ")" ||
-            ch === "[" ||
-            ch === "]") {
-            out.push({ text: ch, raw: ch, start: i, end: i + 1 });
-            i++;
-            continue;
-        }
-        // Bareword: letters, digits, underscore, `.` (schema qualifier), `:` (var
-        // expansion marker is kept inside the word), `$` (positional param).
-        const start = i;
-        while (i < input.length) {
-            const c = input[i];
-            if (c === " " ||
-                c === "\t" ||
-                c === "\n" ||
-                c === "\r" ||
-                c === "," ||
-                c === ";" ||
-                c === "(" ||
-                c === ")" ||
-                c === "[" ||
-                c === "]" ||
-                c === '"' ||
-                c === "'" ||
-                c === "\\") {
-                break;
-            }
-            i++;
-        }
-        out.push({
-            text: input.slice(start, i),
-            raw: input.slice(start, i),
-            start,
-            end: i,
-        });
-    }
-    return out;
+* Split `input` into psql-style words for completion purposes.
+*
+*   - Whitespace is the basic separator.
+*   - "..."           treated as ONE word (quoted identifier).
+*   - '...'           treated as ONE word (string literal).
+*   - \backslash...   treated as ONE word (the backslash command).
+*   - Punctuation (commas, parens, semicolons) is broken into its own word.
+*   - Embedded `.` (schema-qualifier) keeps the dotted name as a single word
+*     so completion sees `pg_catalog.pg_class` whole.
+*
+* The tokenizer is intentionally pragmatic — it doesn't validate SQL, it
+* just slices well enough for the rule body to inspect previous words.
+*/
+const tokenize = (input) => {
+	const out = [];
+	let i = 0;
+	while (i < input.length) {
+		const ch = input[i];
+		if (ch === " " || ch === "	" || ch === "\n" || ch === "\r") {
+			i++;
+			continue;
+		}
+		if (ch === "\\") {
+			const start = i;
+			i++;
+			if (i < input.length && /[!?]/.test(input[i])) i++;
+			else {
+				while (i < input.length && /[A-Za-z_]/.test(input[i])) i++;
+				while (i < input.length && input[i] === "+") i++;
+			}
+			out.push({
+				text: input.slice(start, i),
+				raw: input.slice(start, i),
+				start,
+				end: i
+			});
+			continue;
+		}
+		if (ch === "\"") {
+			const start = i;
+			i++;
+			while (i < input.length) {
+				if (input[i] === "\"") {
+					if (input[i + 1] === "\"") {
+						i += 2;
+						continue;
+					}
+					i++;
+					break;
+				}
+				i++;
+			}
+			out.push({
+				text: input.slice(start, i),
+				raw: input.slice(start, i),
+				start,
+				end: i
+			});
+			continue;
+		}
+		if (ch === "'") {
+			const start = i;
+			i++;
+			while (i < input.length) {
+				if (input[i] === "\\" && i + 1 < input.length) {
+					i += 2;
+					continue;
+				}
+				if (input[i] === "'") {
+					if (input[i + 1] === "'") {
+						i += 2;
+						continue;
+					}
+					i++;
+					break;
+				}
+				i++;
+			}
+			out.push({
+				text: input.slice(start, i),
+				raw: input.slice(start, i),
+				start,
+				end: i
+			});
+			continue;
+		}
+		if (ch === "," || ch === ";" || ch === "(" || ch === ")" || ch === "[" || ch === "]") {
+			out.push({
+				text: ch,
+				raw: ch,
+				start: i,
+				end: i + 1
+			});
+			i++;
+			continue;
+		}
+		const start = i;
+		while (i < input.length) {
+			const c = input[i];
+			if (c === " " || c === "	" || c === "\n" || c === "\r" || c === "," || c === ";" || c === "(" || c === ")" || c === "[" || c === "]" || c === "\"" || c === "'" || c === "\\") break;
+			i++;
+		}
+		out.push({
+			text: input.slice(start, i),
+			raw: input.slice(start, i),
+			start,
+			end: i
+		});
+	}
+	return out;
 };
 /**
- * Slice `input` at `cursor`, then determine what's being completed:
- *
- *   - `prevWords`: the completed words ENTIRELY before the cursor.
- *   - `currentWord`: the partial word the cursor is sitting in (may be '').
- *   - `replaceLength`: number of code points to chop off the input before
- *      inserting the completion (= length of currentWord in code points).
- *
- * The tokenizer is run on `input.slice(0, cursor)` and the LAST token, if
- * it ends exactly at cursor and didn't start with whitespace/punctuation,
- * becomes the current word. Otherwise the current word is empty (the user
- * just typed a space and is starting a new word).
- */
-export const splitForCompletion = (input, cursor) => {
-    const head = input.slice(0, cursor);
-    const tokens = tokenize(head);
-    if (tokens.length === 0) {
-        return { prevWords: [], currentWord: "", replaceLength: 0 };
-    }
-    const last = tokens[tokens.length - 1];
-    // The cursor is sitting inside the last token if (a) it ends exactly at
-    // cursor AND (b) the last char before cursor isn't whitespace.
-    const charBefore = head[head.length - 1];
-    const inWhitespace = charBefore === " " ||
-        charBefore === "\t" ||
-        charBefore === "\n" ||
-        charBefore === "\r";
-    if (last.end === head.length && !inWhitespace) {
-        return {
-            prevWords: tokens.slice(0, -1).map((t) => t.text),
-            currentWord: last.text,
-            replaceLength: Array.from(last.text).length,
-        };
-    }
-    return {
-        prevWords: tokens.map((t) => t.text),
-        currentWord: "",
-        replaceLength: 0,
-    };
+* Slice `input` at `cursor`, then determine what's being completed:
+*
+*   - `prevWords`: the completed words ENTIRELY before the cursor.
+*   - `currentWord`: the partial word the cursor is sitting in (may be '').
+*   - `replaceLength`: number of code points to chop off the input before
+*      inserting the completion (= length of currentWord in code points).
+*
+* The tokenizer is run on `input.slice(0, cursor)` and the LAST token, if
+* it ends exactly at cursor and didn't start with whitespace/punctuation,
+* becomes the current word. Otherwise the current word is empty (the user
+* just typed a space and is starting a new word).
+*/
+const splitForCompletion = (input, cursor) => {
+	const head = input.slice(0, cursor);
+	const tokens = tokenize(head);
+	if (tokens.length === 0) return {
+		prevWords: [],
+		currentWord: "",
+		replaceLength: 0
+	};
+	const last = tokens[tokens.length - 1];
+	const charBefore = head[head.length - 1];
+	const inWhitespace = charBefore === " " || charBefore === "	" || charBefore === "\n" || charBefore === "\r";
+	if (last.end === head.length && !inWhitespace) return {
+		prevWords: tokens.slice(0, -1).map((t) => t.text),
+		currentWord: last.text,
+		replaceLength: Array.from(last.text).length
+	};
+	return {
+		prevWords: tokens.map((t) => t.text),
+		currentWord: "",
+		replaceLength: 0
+	};
 };
+//#endregion
+export { HeadMatches, MatchAny, MatchAnyExcept, Matches, TailMatches, splitForCompletion, tokenize, wordMatches };