From 4a04c491aff43c05b9bbdbe77213ed4e91ad5f1f Mon Sep 17 00:00:00 2001
From: Alan Pierce <alangpierce@gmail.com>
Date: Mon, 27 Jun 2022 14:44:55 -0700
Subject: [PATCH] Fix bug in scientific notation parsing

Fixes #676

Scientific notation parsing was using the shared `readInt` code path, which I
had made more flexible to handle hex digits, but that means that literals like
`1e2` were being treated as a complete number (with the hex digit `e`) rather
than as scientific notation. Normally this wouldn't be a problem since it's all
a number token anyway, but it caused trouble when using dot-style property
access like `1e2.toString()`. Expressions like `1.toString()` are expected to
fail because `.` is interpreted as a decimal point, but when using scientific
notation, the parser needs to be smart enough see that a decimal point wouldn't
be valid, and therefore not include it as part of the token.

The fix was to change `readInt` to only handle decimal digits, not hex digits,
and to inline the hex case into `readRadixNumber`. This should make number
parsing slightly faster because it's not checking for hex digits anymore.
---
 src/parser/tokenizer/index.ts | 40 ++++++++++++++++---------------
 test/sucrase-test.ts          | 44 +++++++++++++++++++++++++++++++++++
 2 files changed, 65 insertions(+), 19 deletions(-)

diff --git a/src/parser/tokenizer/index.ts b/src/parser/tokenizer/index.ts
index 63bdc423..ca62c1aa 100644
--- a/src/parser/tokenizer/index.ts
+++ b/src/parser/tokenizer/index.ts
@@ -751,9 +751,26 @@ function readRegexp(): void {
   finishToken(tt.regexp);
 }
 
-// Read an integer. We allow any valid digit, including hex digits, plus numeric separators, and
-// stop at any other character.
+/**
+ * Read a decimal integer. Note that this can't be unified with the similar code
+ * in readRadixNumber (which also handles hex digits) because "e" needs to be
+ * the end of the integer so that we can properly handle scientific notation.
+ */
 function readInt(): void {
+  while (true) {
+    const code = input.charCodeAt(state.pos);
+    if ((code >= charCodes.digit0 && code <= charCodes.digit9) || code === charCodes.underscore) {
+      state.pos++;
+    } else {
+      break;
+    }
+  }
+}
+
+function readRadixNumber(): void {
+  state.pos += 2; // 0x
+
+  // Walk to the end of the number, allowing hex digits.
   while (true) {
     const code = input.charCodeAt(state.pos);
     if (
@@ -767,29 +784,14 @@ function readInt(): void {
       break;
     }
   }
-}
-
-function readRadixNumber(): void {
-  let isBigInt = false;
-  const start = state.pos;
-
-  state.pos += 2; // 0x
-  readInt();
 
   const nextChar = input.charCodeAt(state.pos);
   if (nextChar === charCodes.lowercaseN) {
     ++state.pos;
-    isBigInt = true;
-  } else if (nextChar === charCodes.lowercaseM) {
-    unexpected("Invalid decimal", start);
-  }
-
-  if (isBigInt) {
     finishToken(tt.bigint);
-    return;
+  } else {
+    finishToken(tt.num);
   }
-
-  finishToken(tt.num);
 }
 
 // Read an integer, octal integer, or floating-point number.
diff --git a/test/sucrase-test.ts b/test/sucrase-test.ts
index 5f41d50c..de25deef 100644
--- a/test/sucrase-test.ts
+++ b/test/sucrase-test.ts
@@ -1469,4 +1469,48 @@ describe("sucrase", () => {
       {transforms: ["typescript"]},
     );
   });
+
+  it("parses scientific notation number literals followed by dot", () => {
+    assertResult(
+      `
+      console.log(1e5.toString());
+    `,
+      `"use strict";
+      console.log(1e5.toString());
+    `,
+    );
+  });
+
+  it("handles parsing of hex literals", () => {
+    assertResult(
+      `
+      const x = 0x8badf00d;
+    `,
+      `"use strict";
+      const x = 0x8badf00d;
+    `,
+    );
+  });
+
+  it("handles parsing of hex bigint literals", () => {
+    assertResult(
+      `
+      const x = 0xabcden;
+    `,
+      `"use strict";
+      const x = 0xabcden;
+    `,
+    );
+  });
+
+  it("handles parsing of negative exponents", () => {
+    assertResult(
+      `
+      const x = 1e-10;
+    `,
+      `"use strict";
+      const x = 1e-10;
+    `,
+    );
+  });
 });