Add a hack for tree-sitter maybe-null-element arrays

Author: tmr232

src/control-flow/cfg-c.ts

@@ -18,6 +18,7 @@ import {
   GenericCFGBuilder,
   type StatementHandlers,
 } from "./generic-cfg-builder.ts";
+import { treeSitterNoNullNodes } from "./hacks.ts";
 import { buildSwitch, collectCases } from "./switch-utils.ts";
 
 function getChildFieldText(node: SyntaxNode, fieldName: string): string {
@@ -94,9 +95,9 @@ const caseTypes = new Set(["case_statement"]);
 
 function getCases(switchSyntax: SyntaxNode): SyntaxNode[] {
   const switchBody = switchSyntax.namedChildren[1] as SyntaxNode;
-  return switchBody.namedChildren
-    .filter((x) => x !== null)
-    .filter((child) => caseTypes.has(child.type));
+  return treeSitterNoNullNodes(switchBody.namedChildren).filter((child) =>
+    caseTypes.has(child.type),
+  );
 }
 
 function parseCase(caseSyntax: SyntaxNode): {
@@ -105,9 +106,9 @@ function parseCase(caseSyntax: SyntaxNode): {
   hasFallthrough: boolean;
 } {
   const isDefault = !caseSyntax.childForFieldName("value");
-  const consequence = caseSyntax.namedChildren
-    .filter((x) => x !== null)
-    .slice(isDefault ? 0 : 1);
+  const consequence = treeSitterNoNullNodes(caseSyntax.namedChildren).slice(
+    isDefault ? 0 : 1,
+  );
   const hasFallthrough = true;
   return { isDefault, consequence, hasFallthrough };
 }

src/control-flow/cfg-go.ts

@@ -16,6 +16,7 @@ import {
   GenericCFGBuilder,
   type StatementHandlers,
 } from "./generic-cfg-builder";
+import { treeSitterNoNullNodes } from "./hacks.ts";
 import { type SwitchOptions, buildSwitch, collectCases } from "./switch-utils";
 
 const processBreakStatement = labeledBreakProcessor(`
@@ -322,9 +323,9 @@ const caseTypes = new Set([
 ]);
 
 function getCases(switchSyntax: SyntaxNode): SyntaxNode[] {
-  return switchSyntax.namedChildren
-    .filter((x) => x !== null)
-    .filter((child) => caseTypes.has(child.type));
+  return treeSitterNoNullNodes(switchSyntax.namedChildren).filter((child) =>
+    caseTypes.has(child.type),
+  );
 }
 
 function parseCase(caseSyntax: SyntaxNode): {
@@ -333,9 +334,9 @@ function parseCase(caseSyntax: SyntaxNode): {
   hasFallthrough: boolean;
 } {
   const isDefault = caseSyntax.type === "default_case";
-  const consequence = caseSyntax.namedChildren
-    .filter((x) => x !== null)
-    .slice(isDefault ? 0 : 1);
+  const consequence = treeSitterNoNullNodes(caseSyntax.namedChildren).slice(
+    isDefault ? 0 : 1,
+  );
   const hasFallthrough = consequence
     .map((node) => node.type)
     .includes("fallthrough_statement");

src/control-flow/cfg-typescript.ts

@@ -20,6 +20,7 @@ import {
   GenericCFGBuilder,
   type StatementHandlers,
 } from "./generic-cfg-builder.ts";
+import { treeSitterNoNullNodes } from "./hacks.ts";
 import { buildSwitch, collectCases } from "./switch-utils.ts";
 
 export function createCFGBuilder(options: BuilderOptions): CFGBuilder {
@@ -115,9 +116,9 @@ const caseTypes = new Set(["switch_case", "switch_default"]);
 
 function getCases(switchSyntax: SyntaxNode): SyntaxNode[] {
   const switchBody = switchSyntax.namedChildren[1] as SyntaxNode;
-  return switchBody.namedChildren
-    .filter((x) => x !== null)
-    .filter((child) => caseTypes.has(child.type));
+  return treeSitterNoNullNodes(switchBody.namedChildren).filter((child) =>
+    caseTypes.has(child.type),
+  );
 }
 
 function parseCase(caseSyntax: SyntaxNode): {
@@ -126,9 +127,9 @@ function parseCase(caseSyntax: SyntaxNode): {
   hasFallthrough: boolean;
 } {
   const isDefault = caseSyntax.type === "switch_default";
-  const consequence = caseSyntax.namedChildren
-    .filter((x) => x !== null)
-    .slice(isDefault ? 0 : 1);
+  const consequence = treeSitterNoNullNodes(caseSyntax.namedChildren).slice(
+    isDefault ? 0 : 1,
+  );
   const hasFallthrough = true;
   return { isDefault, consequence, hasFallthrough };
 }

src/control-flow/common-patterns.ts

@@ -2,6 +2,7 @@ import type { Node as SyntaxNode } from "web-tree-sitter";
 import type { Match } from "./block-matcher.ts";
 import type { BasicBlock } from "./cfg-defs.ts";
 import type { Context } from "./generic-cfg-builder.ts";
+import { treeSitterNoNullNodes } from "./hacks.ts";
 import { last, pairwise, zip } from "./itertools.ts";
 
 export function cStyleIfProcessor(
@@ -514,7 +515,7 @@ export function processStatementSequence(
   ctx: Context,
 ): BasicBlock {
   const blockBlock = ctx.dispatch.many(
-    syntax.namedChildren.filter((x) => x !== null),
+    treeSitterNoNullNodes(syntax.namedChildren),
     syntax,
   );
   ctx.link.syntaxToNode(syntax, blockBlock.entry);

src/control-flow/generic-cfg-builder.ts

@@ -7,6 +7,7 @@ import {
   type BuilderOptions,
   type CFG,
 } from "./cfg-defs";
+import { treeSitterNoNullNodes } from "./hacks.ts";
 import { pairwise } from "./itertools.ts";
 import { NodeMapper } from "./node-mapper";
 
@@ -99,7 +100,7 @@ export class GenericCFGBuilder {
       const blockHandler = new BlockHandler();
       const { entry, exit } = blockHandler.update(
         this.dispatchMany(
-          bodySyntax.namedChildren.filter((x) => x !== null),
+          treeSitterNoNullNodes(bodySyntax.namedChildren),
           bodySyntax,
         ),
       );

src/control-flow/hacks.ts

@@ -0,0 +1,22 @@
+/**
+ * This module contains temporary hacks.
+ * This is terrible code that should not exist,
+ * but is here to circumvent bugs in other projects.
+ *
+ * As such, all code here should document the bugs it works around
+ * and the PRs or issues that will make it obsolete.
+ */
+import type { Node } from "web-tree-sitter";
+
+/**
+ * Assert that there are no nulls in a value returned from tree-sitter.
+ * This can be removed once https://github.com/tree-sitter/tree-sitter/pull/4283 is merged.
+ * Also relates to https://github.com/tree-sitter/tree-sitter/discussions/4273
+ * @param array
+ */
+export function treeSitterNoNullNodes(array: (Node | null)[]): Node[] {
+  if (array.some((x) => x === null)) {
+    throw new Error("tree-sitter API actually had a null in an array!");
+  }
+  return array as Node[];
+}

src/control-flow/overlay.ts

@@ -2,6 +2,7 @@ import type { Container, Element, Svg } from "@svgdotjs/svg.js";
 import type { Node as SyntaxNode } from "web-tree-sitter";
 import type { CFG, GraphNode } from "./cfg-defs.ts";
 import type { AttrMerger } from "./graph-ops.ts";
+import { treeSitterNoNullNodes } from "./hacks.ts";
 import { Lookup } from "./ranges.ts";
 import { svgFromString } from "./svgFromString.ts";
 
@@ -236,7 +237,7 @@ type OverlayRange = {
 };
 
 function parseOverlay(func: SyntaxNode): OverlayRange[] {
-  const comments = func.descendantsOfType("comment").filter((x) => x !== null);
+  const comments = treeSitterNoNullNodes(func.descendantsOfType("comment"));
   const stack: { startOffset: number; text: string }[] = [];
   const overlays: OverlayRange[] = [];
   for (const comment of comments) {