This is a specification for a plain-text, human-friendly format for defining multiple virtual text files in a single physical file, for situations when creating many physical files is undesirable, such as defining test cases for a text format.

Here's a sample HRX that contains two files:

<===> input.scss
ul {
  margin-left: 1em;
  li {
    list-style-type: none;
  }
}

<===> output.css
ul {
  margin-left: 1em;
}
ul li {
  list-style-type: none;
}

HRX files are always encoded in UTF-8.

• Goals
  • Non-Goals
• Syntax
• Semantics
  • Extracting
• Implementations

The HRX format is intended to make it easy to:

• represent multiple chunks of plain-text data in a single physical file;
• read, edit, and create by humans with simple text editors;
• read and modify programatically, while generating easy-to-understand diffs for version control;
• integrate well with tooling for syntax-highlighting or providing other language-specific support for individual files it contains.

The HRX format is not intended as a wire format. The multipart/mixed format defined by MIME is already well-suited to that need.

The HRX format is not intended to faithfully represent any arbitrary directory structure. In order to ensure simplicity and human-readability, it intentionally lacks support for binary data or complex permissions.

The syntax for a HRX file is as follows:

archive        ::= entry* comment?
 
entry          ::= comment? (file | directory)
comment        ::= boundary newline body
file           ::= boundary " "+ path newline body?
directory      ::= boundary " "+ path "/" newline+
boundary       ::= "<" "="+ ">" // must exactly match the first boundary in the archive
newline        ::= U+000A LINE FEED
body           ::= contents newline // no newline at the end of the archive (if the
                                            // archive ends in a body, all trailing
                                            // newlines are part of that body's contents)
contents       ::= any sequence of characters that neither begins with boundary nor
                   includes U+000A LINE FEED followed immediately by boundary
 
path           ::= path-component ("/" path-component)*
path-component ::= path-character+ // not equal to "." or ".."
path-character ::= any character other than U+0000 through U+001F, U+007F DELETE, U+002F
                   SOLIDUS, U+003A COLON, or U+005C REVERSE SOLIDUS

Each path in the file must be codepoint-for-codepoint unique. It's invalid for an initial subsequence of a path's components to be the same as another file's path.

The length of a boundary must be consistent throughout a given HRX file. Longer or shorter boundarys may be included in content blocks, which means a boundary may be chosen for the HRX file that will allow its content blocks to contain arbitrary text.

content blocks don't contain the trailing newline in body. However, the last body in the file doesn't have a trailing newline, so if the file ends with a newline it is included in its content block.

Directories are distinguished from empty files by the trailing "/" in the path name.

A HRX file, represented as an archive, contains a sequence of files and/or directories, each represented as an entry. Each entry has an optional comment that's intended to allow an archive author to include documentation or notes relating to that entry. This comment may contain user-defined metadata that applies to the following file.

Each entry has a path which represents the relative path from the root of the archive to that entry. Each path is divided by U+002F SOLIDUS characters into separate path components. Each component except for the last one represents a nested series of directories; the last component represents the entry itself, which is either a file (for a file) or a directory (for a directory). Parent directories are implicitly assumed to exist if they have entries beneath them, even if they aren't explicitly written as directorys.

Each path component is a string that represents the name of that component. Path component names are interpreted exactly as they appear in the archive file.

File entries with an associated body have their contents given by the body's contents block. File entries with no body have empty contents.

Although HRX files are primarily intended to be used as a replacement for multiple physical files, they can be extracted to the physical filesystem. When a HRX file is extracted, the extraction process should (by default) create a directory named after the HRX file, with the extension ".hrx" removed.

The permissions of extracted files should match those of the original HRX file.

The following packages implement HRX in various languages. If you're writing an implementation, feel free to send a pull request to add it to the list!

• Ruby (original/reference implementation)
• Python
• JavaScript
• Rust
• Java

Tooling for a nicer experience writting HRX files

• VSCode Syntax