unbash - npm Package Compare versions

Comparing version 4.0.6 to 4.0.7

dist/lexer.d.ts

@@ -56,2 +56,3 @@ import type { DeferredCommandExpansion, ParseError, Word, WordPart } from "./types.ts";
     raw: boolean;
+    keywordEligible: boolean;
     constructor(owner?: Lexer | null);
@@ -122,2 +123,3 @@ get value(): string;
     peek(ctx?: LexContext): TokenValue;
+    peekFollow(closers: Uint8Array): TokenValue;
     next(ctx?: LexContext): TokenValue;
@@ -140,2 +142,3 @@ unshift(tok: TokenValue): void;
     private _wordHasExpansions;
+    private _wordKeywordEligible;
     private _wordIsAssignment;
@@ -142,0 +145,0 @@ private _wordAssignmentOperatorPos;

dist/parser.js

@@ -167,2 +167,16 @@ import { hasEmbeddedWordStructure, LexContext, MAX_SYNTAX_NESTING, Token, Lexer, TokenValue } from "./lexer.js";
 listTerminators[Token.DoubleSemiAmp] = 1;
+// After one of these Bash is at a command-start position, the only place a reserved-word
+// terminator may follow with no separator.
+const compoundClosers = new Uint8Array(37);
+compoundClosers[Token.RParen] = 1;
+compoundClosers[Token.RBrace] = 1;
+compoundClosers[Token.DblRBracket] = 1;
+compoundClosers[Token.Fi] = 1;
+compoundClosers[Token.Done] = 1;
+compoundClosers[Token.Esac] = 1;
+compoundClosers[Token.ArithCmd] = 1;
+// Inside `[[ ]]` only an unquoted `!` negates; `'!'` and `\!` are ordinary operands.
+function isTestNegation(t) {
+    return t.token === Token.Word && t.keywordEligible && t.value === "!";
+}
 const commandStarts = new Uint8Array(37);
@@ -272,2 +286,23 @@ commandStarts[Token.Word] = 1;
         const commands = this.list();
+        for (;;) {
+            const unexpected = this.tok.peek(LexContext.CommandStart);
+            if (unexpected.token === Token.EOF)
+                break;
+            this.error(`unexpected token '${unexpected.value}'`, unexpected.pos);
+            // `In` cannot join `listTerminators`: `list()` shares it, and `in` must not terminate a
+            // list inside `for`/`case`.
+            if (!listTerminators[unexpected.token] && unexpected.token !== Token.In)
+                break;
+            this.tok.next(LexContext.CommandStart);
+            let separator = this.tok.peek(LexContext.CommandStart).token;
+            if (separator !== Token.Semi && separator !== Token.Newline && separator !== Token.Amp)
+                break;
+            while (separator === Token.Semi || separator === Token.Newline || separator === Token.Amp) {
+                this.tok.next(LexContext.CommandStart);
+                separator = this.tok.peek(LexContext.CommandStart).token;
+            }
+            const recovered = this.list();
+            for (let i = 0; i < recovered.length; i++)
+                commands.push(recovered[i]);
+        }
         const lexerErrors = this.tok._errors;
@@ -279,2 +314,4 @@ if (lexerErrors !== null && lexerErrors.length > 0) {
         }
+        if (this.errors !== null && this.errors.length > 1)
+            this.errors.sort((a, b) => a.pos - b.pos);
         const result = {
@@ -336,3 +373,3 @@ type: "Script",
         for (;;) {
-            t = this.tok.peek(LexContext.Normal).token;
+            t = this.tok.peekFollow(compoundClosers).token;
             if (t !== Token.Semi && t !== Token.Newline && t !== Token.Amp)
@@ -408,8 +445,8 @@ break;
         let pipelinePos = 0;
-        if (this.tok.peek(LexContext.CommandStart).token === Token.Word &&
-            this.tok.peek(LexContext.CommandStart).value === "time") {
+        const firstToken = this.tok.peek(LexContext.CommandStart);
+        if (firstToken.token === Token.Word && firstToken.keywordEligible && firstToken.value === "time") {
             time = true;
             pipelinePos = this.tok.next(LexContext.CommandStart).pos;
-            if (this.tok.peek(LexContext.CommandStart).token === Token.Word &&
-                this.tok.peek(LexContext.CommandStart).value === "-p")
+            const flag = this.tok.peek(LexContext.CommandStart);
+            if (flag.token === Token.Word && flag.keywordEligible && flag.value === "-p")
                 this.tok.next(LexContext.CommandStart);
@@ -519,3 +556,3 @@ }
         let redirects = EMPTY_REDIRECTS;
-        while (this.tok.peek(LexContext.Normal).token === Token.Redirect) {
+        while (this.tok.peekFollow(compoundClosers).token === Token.Redirect) {
             redirects = this.collectRedirect(redirects, LexContext.Normal);
@@ -972,7 +1009,7 @@ }
         let t = this.tok.peek(LexContext.TestMode);
-        if (t.token !== Token.Word || t.value !== "!")
+        if (!isTestNegation(t))
             return this.parseTestPrimary();
         const firstPos = this.tok.next(LexContext.TestMode).pos;
         t = this.tok.peek(LexContext.TestMode);
-        if (t.token !== Token.Word || t.value !== "!") {
+        if (!isTestNegation(t)) {
             const operand = this.parseTestPrimary();
@@ -982,3 +1019,3 @@ return { type: "TestNot", pos: firstPos, end: operand.end, operand };
         const positions = [firstPos];
-        while (t.token === Token.Word && t.value === "!") {
+        while (isTestNegation(t)) {
             positions.push(this.tok.next(LexContext.TestMode).pos);
@@ -985,0 +1022,0 @@ t = this.tok.peek(LexContext.TestMode);

dist/printer.js

 import { WordImpl } from "./word.js";
 export function print(script) {
-    let out = "";
-    if (script.shebang)
-        out += script.shebang + "\n\n";
-    out += stmts(script.commands, 0);
-    return out;
+    // Saved and restored, not cleared: a lazy getter can re-enter print(), and clearing would
+    // drop the outer call's pending heredocs.
+    const outerQueued = heredocQueue.length;
+    try {
+        let out = "";
+        if (script.shebang)
+            out += script.shebang + "\n\n";
+        out += stmts(script.commands, 0);
+        return out;
+    }
+    finally {
+        heredocQueue.length = outerQueued;
+    }
 }
@@ -23,2 +31,3 @@ function isFunc(s) {
     const pad = "  ".repeat(indent);
+    const queued = heredocQueue.length;
     let out = pad + printNode(s.command, indent);
@@ -29,22 +38,39 @@ for (const r of s.redirects)
         out += " &";
-    out += heredocBodies(s);
-    return out;
+    return heredocQueue.length === queued ? out : flushHeredocs(out, queued);
 }
-function heredocBodies(s) {
+// Bash reads a heredoc body from the line after its `<<`. `redir` marks where it printed and
+// this drains each mark at the end of its own line, so reflowing cannot misplace a body.
+const HEREDOC_MARK = "\u0000";
+const heredocQueue = [];
+function flushHeredocs(text, queued) {
+    if (heredocQueue.length === queued)
+        return text;
+    const pending = heredocQueue.splice(queued);
+    if (pending.length === 1 && text.endsWith(HEREDOC_MARK)) {
+        return text.slice(0, -1) + "\n" + pending[0].content + delimName(pending[0]);
+    }
     let out = "";
-    if (s.command.type === "Command") {
-        for (const r of s.command.redirects)
-            out += heredocBody(r);
+    let copied = 0;
+    let next = 0;
+    let mark = text.indexOf(HEREDOC_MARK);
+    while (mark !== -1) {
+        let lineEnd = text.indexOf("\n", mark);
+        if (lineEnd === -1)
+            lineEnd = text.length;
+        let marks = 0;
+        while (mark !== -1 && mark < lineEnd) {
+            out += text.slice(copied, mark);
+            copied = mark + 1;
+            marks++;
+            mark = text.indexOf(HEREDOC_MARK, copied);
+        }
+        out += text.slice(copied, lineEnd);
+        copied = lineEnd;
+        // `content` already ends in a newline, so this lands the delimiter on its own line.
+        for (; marks > 0 && next < pending.length; marks--, next++) {
+            out += "\n" + pending[next].content + delimName(pending[next]);
+        }
     }
-    for (const r of s.redirects)
-        out += heredocBody(r);
-    return out;
+    return out + text.slice(copied);
 }
-function heredocBody(r) {
-    if (r.operator !== "<<" && r.operator !== "<<-")
-        return "";
-    if (r.content == null)
-        return "";
-    return "\n" + r.content + delimName(r);
-}
 function delimName(r) {
@@ -431,2 +457,6 @@ return r.target?.value ?? "";
     }
+    if (r.content != null && (r.operator === "<<" || r.operator === "<<-")) {
+        heredocQueue.push(r);
+        out += HEREDOC_MARK;
+    }
     return out;
@@ -433,0 +463,0 @@ }

package.json

 {
   "name": "unbash",
-  "version": "4.0.6",
+  "version": "4.0.7",
   "description": "Fast 0-deps bash parser written in TypeScript",
@@ -5,0 +5,0 @@ "keywords": [

README.md

@@ -5,2 +5,4 @@ # unbash
 
+[![NPM Version][2]][1] [![NPM Downloads][3]][1]
+
 ## Install
@@ -52,4 +54,4 @@
 
-To parse `process.argv` (`string[]`), use Node.js [`parseArgs`][1] or a CLI
-library such as [yargs][2] or [citty][3].
+To parse `process.argv` (`string[]`), use Node.js [`parseArgs`][4] or a CLI
+library such as [yargs][5] or [citty][6].
 
@@ -80,2 +82,4 @@ ## Usage
 
+See the full AST at [unbash.statichost.page#input=if \[ -f...][7]
+
 ### Word parts
@@ -126,3 +130,5 @@
 ```js
-const nested = word.parts.find((part) => part.type === "CommandExpansion").script;
+const nested = word.parts.find(
+  (part) => part.type === "CommandExpansion",
+).script;
 const command = nested.commands[0].command;
@@ -166,3 +172,3 @@
 
-[tree-sitter-bash][4] is the right choice when you need:
+[tree-sitter-bash][8] is the right choice when you need:
 
@@ -182,7 +188,7 @@ - Incremental parsing
 - Best-effort error recovery that preserves a partial AST and collects errors
-- Also see [tree-sitter-bash gaps covered by unbash][5]
+- Also see [tree-sitter-bash gaps covered by unbash][9]
 
 ## unbash vs sh-syntax
 
-[sh-syntax][6] is a WASM wrapper around the robust [mvdan/sh][7] Go parser. It
+[sh-syntax][10] is a WASM wrapper around the robust [mvdan/sh][11] Go parser. It
 is highly recommended if you need:
@@ -199,8 +205,8 @@
 - Best-effort partial ASTs for malformed or incomplete editor and user input
-- Also see [sh-syntax gaps covered by unbash][8]
+- Also see [sh-syntax gaps covered by unbash][12]
 
 ## unbash vs bash-parser
 
-[bash-parser][9] (last publish: 2017) and its fork
-[@ericcornelissen/bash-parser][10] (community dependency maintenance fork ❤️ now
+[bash-parser][13] (last publish: 2017) and its fork
+[@ericcornelissen/bash-parser][14] (community dependency maintenance fork ❤️ now
 archived) provide:
@@ -248,4 +254,4 @@
 
-- [unbash.statichost.page][11]
-- [ast-explorer.dev][12]
+- [unbash.statichost.page][15]
+- [ast-explorer.dev][16]
 
@@ -256,13 +262,17 @@ ## License
 
-[1]: https://nodejs.org/api/util.html#utilparseargsconfig
-[2]: https://yargs.js.org/
-[3]: https://www.npmjs.com/package/citty
-[4]: https://github.com/tree-sitter/tree-sitter-bash
-[5]: https://github.com/webpro-nl/unbash/issues/6
-[6]: https://github.com/un-ts/sh-syntax
-[7]: https://github.com/mvdan/sh
-[8]: https://github.com/webpro-nl/unbash/issues/7
-[9]: https://github.com/vorpaljs/bash-parser
-[10]: https://github.com/ericcornelissen/bash-parser
-[11]: https://unbash.statichost.page
-[12]: https://ast-explorer.dev/#eNoVjDsKwzAQRK8yDK5CyAGS2nVAId02jixZAbFr/Kls393rbh7zeBsrn5xLqpV3jr5X/XVzcYgOKRaD8LoNzffTBqFotgkZf8XtUW14oTdRIHaLq00WYscwpRFtCO8g2psm75n3toPHCdz+Ivg=
+[1]: https://www.npmx.dev/package/unbash
+[2]: https://img.shields.io/npm/v/unbash?color=f56e0f
+[3]: https://img.shields.io/npm/dm/unbash?color=f56e0f
+[4]: https://nodejs.org/api/util.html#utilparseargsconfig
+[5]: https://yargs.js.org/
+[6]: https://www.npmjs.com/package/citty
+[7]: https://unbash.statichost.page/#input=if%20%5B%20-f%20%22%241%22%20%5D%3B%20then%20cat%20%22%241%22%3B%20fi
+[8]: https://github.com/tree-sitter/tree-sitter-bash
+[9]: https://github.com/webpro-nl/unbash/issues/6
+[10]: https://github.com/un-ts/sh-syntax
+[11]: https://github.com/mvdan/sh
+[12]: https://github.com/webpro-nl/unbash/issues/7
+[13]: https://github.com/vorpaljs/bash-parser
+[14]: https://github.com/ericcornelissen/bash-parser
+[15]: https://unbash.statichost.page
+[16]: https://ast-explorer.dev/#eNoVjDsKwzAQRK8yDK5CyAGS2nVAId02jixZAbFr/Kls393rbh7zeBsrn5xLqpV3jr5X/XVzcYgOKRaD8LoNzffTBqFotgkZf8XtUW14oTdRIHaLq00WYscwpRFtCO8g2psm75n3toPHCdz+Ivg=

No alert changes

Improved metrics

Total package byte prevSize

265336

2.42%

Lines of code

6802

1.78%

Number of lines in readme file

270

3.85%

No dependency changes