add runtime-specific error handling

1 files changed, 128 insertions, 10 deletions

diff --git a/src/gleam_json_ffi.mjs b/src/gleam_json_ffi.mjs
index 9d3daee..0116ab3 100644
--- a/src/gleam_json_ffi.mjs
+++ b/src/gleam_json_ffi.mjs
@@ -1,6 +1,12 @@
 import { Ok, Error } from './gleam.mjs'
 import { UnexpectedByte, UnexpectedEndOfInput } from './gleam/json.mjs'
 
+const Runtime = {
+  chromium: 'chromium',
+  spidermonkey: 'spidermonkey',
+  jscore: 'jscore',
+}
+
 export function json_to_string(json) {
   return JSON.stringify(json)
 }
@@ -26,29 +32,141 @@ export function decode(string) {
     const result = JSON.parse(string)
     return new Ok(result)
   } catch (err) {
-    return new Error(getJsonDecodeError(err))
+    return new Error(getJsonDecodeError(err, string))
   }
 }
 
-function getJsonDecodeError(stdErr) {
-  if (isUnexpectedByte(stdErr)) return new toUnexpectedByteError(stdErr)
+function getJsonDecodeError(stdErr, json) {
   if (isUnexpectedEndOfInput(stdErr)) return new UnexpectedEndOfInput()
+  const unexpectedByteRuntime = getUnexpectedByteRuntime(stdErr)
+  if (unexpectedByteRuntime) return toUnexpectedByteError(unexpectedByteRuntime, stdErr, json)
   return new UnexpectedByte('', 0)
 }
 
+/**
+ * Matches unexpected end of input messages in:
+ * - Chromium (edge, chrome, node)
+ * - Spidermonkey (firefox)
+ * - JavascriptCore (safari)
+ * 
+ * Note that Spidermonkey and JavascriptCore will both incorrectly report some
+ * UnexpectedByte errors as UnexpectedEndOfInput errors. For example:
+ * 
+ * @example
+ * // in JavascriptCore
+ * JSON.parse('{"a"]: "b"})
+ * // => JSON Parse error: Expected ':' before value
+ * 
+ * JSON.parse('{"a"')
+ * // => JSON Parse error: Expected ':' before value
+ * 
+ * // in Chromium (correct)
+ * JSON.parse('{"a"]: "b"})
+ * // => Unexpected token ] in JSON at position 4
+ * 
+ * JSON.parse('{"a"')
+ * // => Unexpected end of JSON input
+ * 
+ * // in Chromium (correct)
+ * 
+ */
+const unexpectedEndOfInputRegex = /((unexpected (end|eof))|(end of data)|(unterminated string)|(json( parse error|\.parse)\: expected '(\:|\}|\])'))/i
+
 function isUnexpectedEndOfInput(err) {
-  return err.message.includes('Unexpected end') || err.message.includes('Unexpected EOF')
+  return unexpectedEndOfInputRegex.test(err.message)
 }
 
-const unexpectedByteRegex = /Unexpected token (.) in JSON at position (\d+)/
 
-function isUnexpectedByte(err) {
-  return unexpectedByteRegex.test(err)
+/**
+ * Matches unexpected byte messages in:
+ * - Chromium (edge, chrome, node)
+ * 
+ * Matches the character and its position.
+ */
+const chromiumUnexpectedByteRegex = /unexpected token (.) in JSON at position (\d+)/
+
+/**
+ * Matches unexpected byte messages in:
+ * - JavascriptCore (safari)
+ * 
+ * JavascriptCore only reports what the character is and not its position.
+ */
+const jsCoreUnexpectedByteRegex = /unexpected identifier "(.)"/i
+
+/**
+ * Matches unexpected byte messages in:
+ * - Spidermonkey (firefox)
+ * 
+ * Matches the position in a 2d grid only and not the character.
+ */
+const spidermonkeyUnexpectedByteRegex = /((unexpected character|expected double-quoted property name) at line (\d+) column (\d+))/i
+
+function getUnexpectedByteRuntime(err) {
+  if (chromiumUnexpectedByteRegex.test(err.message)) return Runtime.chromium
+  if (jsCoreUnexpectedByteRegex.test(err.message)) return Runtime.jscore
+  if (spidermonkeyUnexpectedByteRegex.test(err.message)) return Runtime.spidermonkey
+  return null
+}
+
+function toUnexpectedByteError(runtime, err, json) {
+  switch (runtime) {
+    case Runtime.chromium:
+      return toChromiumUnexpectedByteError(err)
+    case Runtime.spidermonkey:
+      return toSpidermonkeyUnexpectedByteError(err, json)
+    case Runtime.jscore:
+      return toJsCoreUnexpectedByteError(err)
+  }
 }
 
-function toUnexpectedByteError(err) {
-  const match = unexpectedByteRegex.exec(err.message)
-  const byte = "0x" + match[1].charCodeAt(0).toString(16).toUpperCase()
+function toChromiumUnexpectedByteError(err) {
+  const match = chromiumUnexpectedByteRegex.exec(err.message)
+  const byte = toHex(match[1])
   const position = Number(match[2])
   return new UnexpectedByte(byte, position)
 }
+
+function toSpidermonkeyUnexpectedByteError(err, json) {
+  const match = spidermonkeyUnexpectedByteRegex.exec(err.message)
+  const line = Number(match[1])
+  const column = Number(match[2])
+  const position = getPositionFromMultiline(line, column, json)
+  const byte = toHex(err.message[position])
+  return new UnexpectedByte(byte, position)
+}
+
+function toJsCoreUnexpectedByteError(err) {
+  const match = jsCoreUnexpectedByteRegex.exec(err.message)
+  const byte = toHex(match[1])
+  return new UnexpectedByte(byte, 0)
+}
+
+function toHex(char) {
+  return "0x" + char.charCodeAt(0).toString(16).toUpperCase()
+}
+
+/**
+ * Gets the position of a character in a flattened (i.e. single line) string
+ * from a line and column number. Note that the position is 0-indexed and 
+ * the line and column numbers are 1-indexed.
+ * 
+ * @param {number} line 
+ * @param {number} column 
+ * @param {string} string 
+ */
+function getPositionFromMultiline(line, column, string) {
+  if (line === 1) return column - 1
+
+  let currentLn = 1
+  let position = 0
+  string.split('').find((char, idx) => {
+    if (char === '\n') currentLn += 1
+    if (currentLn === line) {
+      position = idx + column
+      return true
+    }
+    return false
+  })
+
+  return position
+}
\ No newline at end of file