Source Map Format Specification

Introduction

This Ecma Standard defines the Source Map Format, used for mapping transpiled source code back to the original sources.

The source map format has the following goals:

• Support source-level debugging allowing bidirectional mapping

• Support server-side stack trace deobfuscation

The document at https://tc39.es/source-map/ is the most accurate and up-to-date source map specification. It contains the content of the most recently published snapshot plus any modifications that will be included in the next snapshot.

If you want to get involved you will find more information at the specification repository.

The original source map format (v1) was created by Joseph Schorr for use by Closure Inspector to enable source-level debugging of optimized JavaScript code (although the format itself is language agnostic). However, as the size of the projects using source maps expanded, the verbosity of the format started to become a problem. The v2 format [V2Format] was created by trading some simplicity and flexibility to reduce the overall size of the source map. Even with the changes made with the v2 version of the format, the source map file size was limiting its usefulness. The v3 format is based on suggestions made by Pavel Podivilov (Google).

The source map format does not have version numbers anymore, and it is instead hard-coded to always be "3".

1. Scope

This Standard defines the source map format, used by different types of developer tools to improve the debugging experience of code compiled to JavaScript, WebAssembly, and CSS.

2. Conformance

A conforming source map document is a JSON document that conforms to the structure detailed in this specification.

A conforming source map generator should generate documents which are conforming source map documents, and can be decoded by the algorithms in this specification without reporting any errors (even those which are specified as optional).

A conforming source map consumer should implement the algorithms specified in this specification for retrieving (where applicable) and decoding source map documents. A conforming consumer is permitted to ignore errors or report them without terminating where the specification indicates that an algorithm may optionally report an error.

3. References

Normative References

4. Terms and definitions

For the purposes of this document, the following terms and definitions apply.

Generated Code

The code which is generated by the compiler or transpiler.

Original Source

The source code which has not been passed through the compiler.

Base64 VLQ [VLQ]

A [base64] value, where the most significant bit (the 6th bit) is used as the continuation bit, and the "digits" are encoded into the string least significant first, and where the least significant bit of the first digit is used as the sign bit.

Note: The values that can be represented by the VLQ Base64 encoded are limited to 32-bit quantities until some use case for larger values is presented. This means that values exceeding 32-bits are invalid and implementations may reject them. The sign bit is counted towards the limit, but the continuation bits are not.

The string "iB" represents a Base64 VLQ with two digits. The first digit "i" encodes the bit pattern 0x100010, which has a continuation bit of 1 (the VLQ continues), a sign bit of 0 (non-negative), and the value bits 0x0001. The second digit B encodes the bit pattern 0x000001, which has a continuation bit of 0, no sign bit, and value bits 0x00001. The decoding of this VLQ string is the number 17.

The string "V" represents a Base64 VLQ with one digit. The digit "V" encodes the bit pattern 0x010101, which has a continuation bit of 0 (no continuation), a sign bit of 1 (negative), and the value bits 0x1010. The decoding of this VLQ string is the number -10.

Source Mapping URL

The URL referencing the location of a source map from the Generated code.

Column

The zero-based indexed offset within a line of the generated code. How this offset is measured can depend on the content type. For JavaScript and CSS based source maps, they are defined to be in UTF-16 code units analogous to JavaScript string indices. That means that "A" (LATIN CAPITAL LETTER A) measures as 1 code unit, and "🔥" (FIRE) measures as 2 code units. For WebAssembly, columns are defined as byte offsets from the beginning of the binary content (and there is only one group representing a line). Source maps for other content types may diverge from this.

5. Source Map Format

A source map is a JSON document containing a top-level JSON object with the following structure:

{
  "version" : 3,
  "file": "out.js",
  "sourceRoot": "",
  "sources": ["foo.js", "bar.js"],
  "sourcesContent": [null, null],
  "names": ["src", "maps", "are", "fun"],
  "mappings": "A,AAAB;;ABCDE"
  "ignoreList": [0]
}

• version is the version field which must always be the number 3 as an integer. The source map may be rejected if the field has any other value.

• file is an optional name of the generated code that this source map is associated with. It’s not specified if this can be a URL, relative path name, or just a base name. Source map generators may choose the appropriate interpretation for their contexts of use.

• sourceRoot is an optional source root string, used for relocating source files on a server or removing repeated values in the sources entry. This value is prepended to the individual entries in the sources field.

• sources is a list of original sources used by the mappings entry. Each entry is either a string that is a (potentially relative) URL or null if the source name is not known.

• sourcesContent is an optional list of source content (i.e., the Original Source) strings, used when the source cannot be hosted. The contents are listed in the same order as the sources. Entries may be null if some original sources should be retrieved by name.

• names is an optional list of symbol names which may be used by the mappings entry.

• mappings is a string with the encoded mapping data (see § 5.1 Mappings Structure).

• ignoreList is an optional list of indices of files that should be considered third party code, such as framework code or bundler-generated code. This allows developer tools to avoid code that developers likely don’t want to see or step through, without requiring developers to configure this beforehand. It refers to the sources array and lists the indices of all the known third-party sources in the source map. Some browsers may also use the deprecated x_google_ignoreList field if ignoreList is not present.

A decoded source map is a struct with the following fields:

file

A string or null.

sources

A list of decoded sources.

mappings

A list of decoded mappings.

A decoded source is a struct with the following fields:

URL

A URL or null.

content

A string or null.

ignored

A boolean.

To decode a source map from a JSON string str given a URL baseURL, run the following steps:

1. Let jsonMap be the result of parsing a JSON string to an Infra value str.

2. If jsonMap is not a map, report an error and abort these steps.

3. Decode a source map given jsonMap and baseURL, and return its result if any.

To decode a source map given a string-keyed map jsonMap and a URL baseURL, run the following steps:

1. If jsonMap["version"] does not exist or jsonMap["version"] is not 3, optionally report an error.

2. If jsonMap["mappings"] does not exist or jsonMap["mappings"], is not a string, throw an error.

3. If jsonMap["sources"] does not exist or jsonMap["sources"], is not a list, throw an error.

4. Let sourceMap be a new decoded source map.

5. Set sourceMap’s file to optionally get a string "file" from jsonMap.

6. Set sourceMap’s sources to the result of decoding source map sources given baseURL with:

   • sourceRoot set to optionally get a string "sourceRoot" from jsonMap;

   • sources set to optionally get a list of optional strings "sources" from jsonMap;

   • sourcesContent set to optionally get a list of optional strings "sourcesContent" from jsonMap;

   • ignoredSources set to optionally get a list of array indexes "ignoreList" from jsonMap.

7. Set sourceMap’s mappings to the result of decoding source map mappings with:

   • mappings set to jsonMap["mappings"];

   • names set to optionally get a list of strings "names" from jsonMap;

   • sources set to sourceMap’s sources.

8. Return sourceMap.

To optionally get a string key from a string-keyed map jsonMap, run the following steps:

1. If jsonMap[key] does not exist, return null.

2. If jsonMap[key] is not a string, optionally report an error and return null.

3. Return jsonMap[key].

To optionally get a list of strings key from a string-keyed map jsonMap, run the following steps:

1. If jsonMap[key] does not exist, return a new empty list.

2. If jsonMap[key] is not a list, optionally report an error and return a new empty list.

3. Let list be a new empty list.

4. For each jsonItem of jsonMap[key]:

   1. If jsonItem is a string, append it to list.

   2. Else, optionally report an error and append "" to list.

5. Return list.

To optionally get a list of optional strings key from a string-keyed map jsonMap, run the following steps:

1. If jsonMap[key] does not exist, return a new empty list.

2. If jsonMap[key] is not a list, optionally report an error and return a new empty list.

3. Let list be a new empty list.

4. For each jsonItem of jsonMap[key]:

   1. If jsonItem is a string, append it to list.

   2. Else,

      1. If jsonItem is not null, optionally report an error.

      2. Append null to list.

5. Return list.

To optionally get a list of array indexes key from a string-keyed map jsonMap, run the following steps:

1. If jsonMap[key] does not exist, return a new empty list.

2. If jsonMap[key] is not a list, optionally report an error and return a new empty list.

3. Let list be a new empty list.

4. For each jsonItem of jsonMap[key]:

   1. If jsonItem is a non-negative integer number, append it to list.

   2. Else,

      1. If jsonItem is not null, optionally report an error.

      2. Append null to list.

5. Return list.

To optionally report an error, implementations can choose to:

• Do nothing.

• Report an error to the user, and continue processing.

• Throw an error to abort the running algorithm. (Infra § 3.6 Control flow)

5.1. Mappings Structure

The mappings data is broken down as follows:

• each group representing a line in the generated file is separated by a semicolon (;)

• each segment is separated by a comma (,)

• each segment is made up of 1, 4, or 5 variable length fields.

The fields in each segment are:

1. The zero-based starting column of the line in the generated code that the segment represents. If this is the first field of the first segment, or the first segment following a new generated line (;), then this field holds the whole Base64 VLQ. Otherwise, this field contains a Base64 VLQ that is relative to the previous occurrence of this field. Note that this is different from the subsequent fields below because the previous value is reset after every generated line.

2. If present, the zero-based index into the sources list. This field contains a Base64 VLQ relative to the previous occurrence of this field, unless it is the first occurrence of this field, in which case the whole value is represented.

3. If present, the zero-based starting line in the original source. This field contains a Base64 VLQ relative to the previous occurrence of this field, unless it is the first occurrence of this field, in which case the whole value is represented. Must be present if there is a source field.

4. If present, the zero-based starting column of the line in the original source. This field contains a Base64 VLQ relative to the previous occurrence of this field, unless it is the first occurrence of this field, in which case the whole value is represented. Must be present if there is a source field.

5. If present, the zero-based index into the names list associated with this segment. This field contains a Base64 VLQ relative to the previous occurrence of this field, unless it is the first occurrence of this field, in which case the whole value is represented.

Note: The purpose of this encoding is to reduce the source map size. VLQ encoding reduced source maps by 50% relative to the [V2Format] in tests performed using Google Calendar.

Note: Segments with one field are intended to represent generated code that is unmapped because there is no corresponding original source code, such as code that is generated by a compiler. Segments with four fields represent mapped code where a corresponding name does not exist. Segments with five fields represent mapped code that also has a mapped name.

Note: Using file offsets was considered but rejected in favor of using line/column data to avoid becoming misaligned with the original due to platform-specific line endings.

A decoded mapping is a struct with the following fields:

generatedLine

A non-negative integer.

generatedColumn

A non-negative integer.

originalSource

A decoded source or null.

originalLine

A non-negative integer or null.

originalColumn

A non-negative integer or null.

name

A string or null.

To decode source map mappings given a string mappings, a list of strings names, and a list of decoded sources sources, run the following steps:

1. Validate base64 VLQ groupings with mappings.

2. Let decodedMappings be a new empty list.

3. Let groups be the result of strictly splitting mappings on ;.

4. Let generatedLine be 0.

5. Let originalLine be 0.

6. Let originalColumn be 0.

7. Let sourceIndex be 0.

8. Let nameIndex be 0.

9. While generatedLine is less than groups’s size:

   1. If groups[generatedLine] is not the empty string, then:

      1. Let segments be the result of strictly splitting groups[generatedLine] on ,.

      2. Let generatedColumn be 0.

      3. For each segment in segments:

         1. Let position be a position variable for segment, initially pointing at segment’s start.

         2. Decode a base64 VLQ from segment given position and let relativeGeneratedColumn be the result.

         3. If relativeGeneratedColumn is null, optionally report an error and continue with the next iteration.

         4. Increase generatedColumn by relativeGeneratedColumn. If the result is negative, optionally report an error and continue with the next iteration.

         5. Let decodedMapping be a new decoded mapping whose generatedLine is generatedLine, generatedColumn is generatedColumn, originalSource is null, originalLine is null, originalColumn is null, and name is null.

         6. Append decodedMapping to decodedMappings.

         7. Decode a base64 VLQ from segment given position and let relativeSourceIndex be the result.

         8. Decode a base64 VLQ from segment given position and let relativeOriginalLine be the result.

         9. Decode a base64 VLQ from segment given position and let relativeOriginalColumn be the result.

         10. If relativeOriginalColumn is null, then:

             1. If relativeSourceIndex is not null, optionally report an error.

             2. Continue with the next iteration.

         11. Increase sourceIndex by relativeSourceIndex.

         12. Increase originalLine by relativeOriginalLine.

         13. Increase originalColumn by relativeOriginalColumn.

         14. If any of sourceIndex, originalLine, or originalColumn are less than 0, or if sourceIndex is greater than or equal to sources’s size, optionally report an error.

         15. Else,

             1. Set decodedMapping’s originalSource to sources[sourceIndex].

             2. Set decodedMapping’s originalLine to originalLine.

             3. Set decodedMapping’s originalColumn to originalColumn.

         16. Decode a base64 VLQ from segment given position and let relativeNameIndex be the result.

         17. If relativeNameIndex is not null, then:

             1. Increase nameIndex by relativeNameIndex.

             2. If nameIndex is negative or greater than or equal to names’s size, optionally report an error.

             3. Else, set decodedMapping’s name to names[nameIndex].

         18. If position does not point to the end of segment, optionally report an error.

   2. Increase generatedLine by 1.

10. Return decodedMappings.

To validate base64 VLQ groupings from a string groupings, run the following steps:

1. If groupings is not an ASCII string, throw an error.

2. If groupings contains any code unit other than:

   • U+002C (,) or U+003B (;);

   • U+0030 (0) to U+0039 (9);

   • U+0041 (A) to U+005A (Z);

   • U+0061 (a) to U+007A (z);

   • U+002B (+), U+002F (/)

   NOTE: These are the valid [base64] characters (excluding the padding character =), together with , and ;.

   then throw an error.

To decode a base64 VLQ from a string segment given a position variable position, run the following steps:

1. If position points to the end of segment, return null.

2. Let first be a byte whose the value is the number corresponding to segment’s positionth code unit, according to the [base64] encoding.

   NOTE: The two most significant bits of first are 0.

3. Let sign be 1 if first & 0x01 is 0x00, and -1 otherwise.

4. Let value be (first >> 1) & 0x0F, as a number.

5. Let nextShift be 16.

6. Let currentByte be first.

7. While currentByte & 0x20 is 0x20:

   1. Advance position by 1.

   2. If position points to the end of segment, throw an error.

   3. Set currentByte to the byte whose the value is the number corresponding to segment’s positionth code unit, according to the [base64] encoding.

   4. Let chunk be currentByte & 0x1F, as a number.

   5. Add chunk * nextShift to value.

   6. If value is greater than or equal to 2³¹, throw an error.

   7. Multiply nextShift by 32.

8. Advance position by 1.

9. If value is 0 and sign is -1, return -2147483648.

   NOTE: -2147483648 is the smallest 32-bit signed integer.

10. Return value * sign.

NOTE: In addition to returning the decoded value, this algorithm updates the position variable in the calling algorithm.

5.1.1. Mappings for generated JavaScript code

Generated code positions that may have mapping entries are defined in terms of input elements as per ECMAScript Lexical Grammar. Mapping entries must point to either:

1. the first code point of the source text matched by IdentifierName, PrivateIdentifier, Punctuator, DivPunctuator, RightBracePunctuator, NumericLiteral and RegularExpressionLiteral.

2. any code point of the source text matched by Comment, HashbangComment, StringLiteral, Template, TemplateSubstitutionTail, WhiteSpace and LineTerminator.

5.1.2. Names for generated JavaScript code

Source map generators should create a mapping entry with a name field for a JavaScript token, if

1. The original source language construct maps semantically to the generated JavaScript code.

2. The original source language construct has a name.

Then the name of the mapping entry should be the name of the original source language construct. A mapping with a non-null name is called a named mapping.

Example: A minifier renaming functions and variables or removing function names from immediately invoked function expressions.

The following enumeration lists productions of the ECMAScript Syntactic Grammar and the respective token or non-terminal (on the right-hand side of the production) for which source map generators should emit a named mapping. The mapping entry created for such tokens must follow § 5.1.1 Mappings for generated JavaScript code.

The enumeration should be understood as the "minimum". In general, source map generators are free to emit any additional named mappings.

Note: The enumeration also lists tokens where generators "may" emit named mappings in addition to the tokens where they "should". These reflect the reality where existing tooling emits or expects named mappings. The duplicated named mapping is comparably cheap: Indices into names are encoded relative to each other so subsequent mappings to the same name are encoded as 0 (A).

1. The BindingIdentifier(s) for LexicalDeclaration, VariableStatement and Parameter List.

2. The BindingIdentifier for FunctionDeclaration, FunctionExpression, AsyncFunctionDeclaration, AsyncFunctionExpression, GeneratorDeclaration, GeneratorExpression, AsyncGeneratorDeclaration, and AsyncGeneratorExpression if it exists, or the opening parenthesis ( preceding the FormalParameters otherwise.

   Source map generators may chose to emit a named mapping on the opening parenthesis regardless of the presence of the BindingIdentifier.

3. For an ArrowFunction or AsyncArrowFunction:

   1. The => token where ArrowFunction is produced with a single BindingIdentifier for ArrowParameters. Or AsyncArrowFunction is produced with an AsyncArrowBindingIdentifier.

      Note: This describes the case of (async) arrow functions with a single parameter, where that single parameter is not wrapped in parenthesis.

   2. The opening parenthesis ( where ArrowFunction or AsyncArrowFunction is produced with ArrowFormalParameters.

      Source map generators may chose to additionally emit a named mapping on the => token for consistency with (1).

4. The ClassElementName for MethodDefinition. This includes generators, async methods, async generators and accessors. For MethodDefinition where ClassElementName is "constructor", the name should be the original class name if applicable.

   Source map generators may chose to additionally emit a named mapping on the opening parenthesis (.

5. Source map generators may emit named mappings for IdentifierReference in Expression.

5.2. Resolving Sources

If the sources are not absolute URLs after prepending the sourceRoot, the sources are resolved relative to the SourceMap (like resolving the script src attribute in an HTML document).

To decode source map sources given a URL baseURL, a string or null sourceRoot, a list of either strings or nulls sources, a list of either strings or nulls sourcesContent, and a list of numbers ignoredSources, run the following steps:

1. Let decodedSources be a new empty list.

2. Let sourceURLPrefix be "".

3. If sourceRoot is not null, then:

   1. If sourceRoot contains the code point U+002F (/), then:

      1. Let index be the index of the last occurrence of U+002F (/) in sourceRoot.

      2. Set sourceURLPrefix to the substring of sourceRoot from 0 to index + 1.

   2. Else, set sourceURLPrefix to the concatenation of sourceRoot and "/".

4. For each source of sources with index index:

   1. Let decodedSource be a new decoded source whose URL is null, content is null, and ignored is false.

   2. If source is not null:

      1. Set source to the concatenation of sourceURLPrefix and source.

      2. Let sourceURL be the result of URL parsing source with baseURL.

      3. If sourceURL is failure, optionally report an error.

      4. Else, set decodedSource’s URL to sourceURL.

   3. If index is in ignoredSources, set decodedSource’s ignored to true.

   4. If sourcesContent’s size is greater than or equal to index, set decodedSource’s content to sourcesContent[index].

   5. Append decodedSource to decodedSources.

5. Return decodedSources.

NOTE: Implementations that support showing source contents but do not support showing multiple sources with the same URL and different content will arbitrarily choose one of the various contents corresponding to the given URL.

5.3. Extensions

Source map consumers must ignore any additional unrecognized properties, rather than causing the source map to be rejected, so that additional features can be added to this format without breaking existing users.

6. Index Map

To support concatenating generated code and other common post-processing, an alternate representation of a map is supported:

{
  "version" : 3,
  "file": "app.js",
  "sections": [
    {
      "offset": {"line": 0, "column": 0},
      "map": {
        "version" : 3,
        "file": "section.js",
        "sources": ["foo.js", "bar.js"],
        "names": ["src", "maps", "are", "fun"],
        "mappings": "AAAA,E;;ABCDE"
      }
    },
    {
      "offset": {"line": 100, "column": 10},
      "map": {
        "version" : 3,
        "file": "another_section.js",
        "sources": ["more.js"],
        "names": ["more", "is", "better"],
        "mappings": "AAAA,E;AACA,C;ABCDE"
      }
    }
  ]
}

The index map follows the form of the standard map. Like the regular source map, the file format is JSON with a top-level object. It shares the version and file field from the regular source map, but gains a new sections field.

sections is an array of Section objects.

6.1. Section

Section objects have the following fields:

• offset is an object with two fields, line and column, that represent the offset into generated code that the referenced source map represents.

• map is an embedded complete source map object. An embedded map does not inherit any values from the containing index map.

The sections must be sorted by starting position and the represented sections must not overlap.

7. Retrieving Source Maps

7.1. Linking generated code to source maps

While the source map format is intended to be language and platform agnostic, it is useful to define how to reference to them for the expected use-case of web server-hosted JavaScript.

There are two possible ways to link source maps to the output. The first requires server support in order to add an HTTP header and the second requires an annotation in the source.

Source maps are linked through URLs as defined in [URL]; in particular, characters outside the set permitted to appear in URIs must be percent-encoded and it may be a data URI. Using a data URI along with sourcesContent allows for a completely self-contained source map.

The HTTP sourcemap header has precedence over a source annotation, and if both are present, the header URL should be used to resolve the source map file.

Regardless of the method used to retrieve the Source Mapping URL the same process is used to resolve it, which is as follows:

When the Source Mapping URL is not absolute, then it is relative to the generated code’s source origin. The source origin is determined by one of the following cases:

• If the generated source is not associated with a script element that has a src attribute and there exists a //# sourceURL comment in the generated code, that comment should be used to determine the source origin.

  Note: Previously, this was //@ sourceURL, as with //@ sourceMappingURL, it is reasonable to accept both but //# is preferred.

• If the generated code is associated with a script element and the script element has a src attribute, the src attribute of the script element will be the source origin.

• If the generated code is associated with a script element and the script element does not have a src attribute, then the source origin will be the page’s origin.

• If the generated code is being evaluated as a string with the eval() function or via new Function(), then the source origin will be the page’s origin.

7.1.1. Linking through HTTP headers

If a file is served through HTTP(S) with a sourcemap header, the value of the header is the URL of the linked source map.

sourcemap: <url>

Note: Previous revisions of this document recommended a header name of x-sourcemap. This is now deprecated; sourcemap is now expected.

7.1.2. Linking through inline annotations

The generated code should include a comment, or the equivalent construct depending on its language or format, named sourceMappingURL and that contains the URL of the source map. This specification defines how the comment should look like for JavaScript, CSS, and WebAssembly. Other languages should follow a similar convention.

For a given language there can be multiple ways of detecting the sourceMappingURL comment, to allow for different implementations to choose what is less complex for them. The generated code unambiguously links to a source map if the result of all the extraction methods is the same.

If a tool consumes one or more source files that unambiguously links to a source map and it produces an output file that links to a source map, it must do so unambiguously.

let a = `
//# sourceMappingURL=foo.js.map
//`;

7.1.2.1. Extraction methods for JavaScript sources

To extract a Source Map URL from JavaScript through parsing a string source, run the following steps:

1. Let tokens be the list of tokens obtained by parsing source according to [ECMA-262].

2. For each token in tokens, in reverse order:

   1. If token is not a single-line comment or a multi-line comment, return null.

   2. Let comment be the content of token.

   3. If matching a Source Map URL in comment returns a string, return it.

3. Return null.

To extract a Source Map URL from JavaScript without parsing a string source, run the following steps:

1. Let lines be the result of strictly splitting source on ECMAScript line terminator code points.

2. Let lastURL be null.

3. For each line in lines:

   1. Let position be a position variable for line, initially pointing at the start of line.

   2. While position doesn’t point past the end of line:

      1. Collect a sequence of code points that are ECMAScript white space code points from line given position.

         NOTE: The collected code points are not used, but position is still updated.

      2. If position points past the end of line, break.

      3. Let first be the code point of line at position.

      4. Increment position by 1.

      5. If first is U+002F (/) and position does not point past the end of line, then:

         1. Let second be the code point of line at position.

         2. Increment position by 1.

         3. If second is U+002F (/), then:

            1. Let comment be the code point substring from position to the end of line.

            2. If matching a Source Map URL in comment returns a string, set lastURL to it.

            3. Break.

         4. Else if second is U+002A (*), then:

            1. Let comment be the empty string.

            2. While position + 1 doesn’t point past the end of line:

               1. Let c1 be the code point of line at position.

               2. Increment position by 1.

               3. Let c2 be the code point of line at position.

               4. If c1 is U+002A (*) and c2 is U+002F (/), then:

                  1. If matching a Source Map URL in comment returns a string, set lastURL to it.

                  2. Increment position by 1.

               5. Append c1 to comment.

         5. Else, set lastURL to null.

      6. Else, set lastURL to null.

      Note: We reset lastURL to null whenever we find a non-comment code character.

4. Return lastURL.

NOTE: The algorithm above has been designed so that the source lines can be iterated in reverse order, returning early after scanning through a line that contains a sourceMappingURL comment.

const JS_NEWLINE = /^/m;

// This RegExp will always match one of the following:
// - single-line comments
// - "single-line" multi-line comments
// - unclosed multi-line comments
// - just trailing whitespaces
// - a code character
// The loop below differentiates between all these cases.
const JS_COMMENT =
  /\s*(?:\/\/(?<single>.*)|\/\*(?<multi>.*?)\*\/|\/\*.*|$|(?<code>[^\/]+))/uym;

const PATTERN = /^[@#]\s*sourceMappingURL=(\S*?)\s*$/;

let lastURL = null;
for (const line of source.split(JS_NEWLINE)) {
  JS_COMMENT.lastIndex = 0;
  while (JS_COMMENT.lastIndex < line.length) {
    let commentMatch = JS_COMMENT.exec(line).groups;
    let comment = commentMatch.single ?? commentMatch.multi;
    if (comment != null) {
      let match = PATTERN.exec(comment);
      if (match !== null) lastURL = match[1];
    } else if (commentMatch.code != null) {
      lastURL = null;
    } else {
      // We found either trailing whitespaces or an unclosed comment.
      // Assert: JS_COMMENT.lastIndex === line.length
    }
  }
}
return lastURL;

To match a Source Map URL in a comment comment (a string), run the following steps:

1. Let pattern be the regular expression /^[@#]\s*sourceMappingURL=(\S*?)\s*$/.

2. Let match be ! RegExpBuiltInExec(pattern, comment).

3. If match is not null, return match[1].

4. Return null.

Note: The prefix for this annotation was initially //@ however this conflicts with Internet Explorer’s Conditional Compilation and was changed to //#.

Source map generators must only emit //# while source map consumers must accept both //@ and //#.

7.1.2.2. Extraction methods for CSS sources

Extracting source mapping URLs from CSS is similar to JavaScript, with the exception that CSS only supports /* ... */-style comments.

7.1.2.3. Extraction methods for WebAssembly binaries

To extract a Source Map URL from a WebAssembly source given a byte sequence bytes, run the following steps:

1. Let module be module_decode(bytes).

2. If module is error, return null.

3. For each custom section customSection of module,

   1. Let name be the name of customSection, decoded as UTF-8.

   2. If name is "sourceMappingURL", then:

      1. Let value be the bytes of customSection, decoded as UTF-8.

      2. If value is failure, return null.

      3. Return value.

4. Return null.

Since WebAssembly is not a textual format and it does not support comments, it supports a single unambiguous extraction method. The URL is encoded using [WasmNamesBinaryFormat], and it’s placed as the content of the custom section. It is invalid for tools that generate WebAssembly code to generate two or more custom sections with the "sourceMappingURL" name.

7.2. Fetching Source Maps

To fetch a source map given a URL url, run the following steps:

1. Let promise be a new promise.

2. Let request be a new request whose URL is url.

3. Fetch request with processResponseConsumeBody set to the following steps given response response and null, failure, or a byte sequence bodyBytes:

   1. If bodyBytes is null or failure, reject promise with a TypeError and abort these steps.

   2. If url’s scheme is an HTTP(S) scheme and bodyBytes starts with `)]}'`, then:

      1. While bodyBytes’s length is not 0 and bodyBytes’s 0th byte is not an HTTP newline byte:

         1. remove the 0th byte from bodyBytes.

         Note: For historic reasons, when delivering source maps over HTTP(S), servers may prepend a line starting with the string )]}' to the source map.

         )]}'garbage here
         {"version": 3, ...}
         

         is interpreted as

         {"version": 3, ...}
         

   3. Let sourceMap be the result of parsing JSON bytes to a JavaScript value given bodyBytes.

   4. If the previous step threw an error, reject promise with that error.

   5. Otherwise, resolve promise with sourceMap.

4. Return promise.

8. Conventions

The following conventions should be followed when working with source maps or when generating them.

8.1. Source Map Naming

Commonly, a source map will have the same name as the generated file but with a .map extension. For example, for page.js a source map named page.js.map would be generated.

8.2. Linking eval’d code to named generated code

There is an existing convention that should be supported for the use of source maps with eval’d code, it has the following form:

//# sourceURL=foo.js

It is described in [EvalSourceURL].

9. Language Neutral Stack Mapping Notes

Stack tracing mapping without knowledge of the source language is not covered by this document.

10. Multi-level Mapping Notes

It is getting more common to have tools generate sources from some DSL (templates) or compile TypeScript -> JavaScript -> minified JavaScript, resulting in multiple translations before the final source map is created. This problem can be handled in one of two ways. The easy but lossy way is to ignore the intermediate steps in the process for the purposes of debugging, the source location information from the translation is either ignored (the intermediate translation is considered the “Original Source”) or the source location information is carried through (the intermediate translation hidden). The more complete way is to support multiple levels of mapping: if the Original Source also has a source map reference, the user is given the choice of using that as well.

However, It is unclear what a "source map reference" looks like in anything other than JavaScript. More specifically, what a source map reference looks like in a language that doesn’t support JavaScript-style single-line comments.

11. License

This work is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License.