This document is detailed notes on the Pod markup language. Most people will only have to read perlpod to know how to write in Pod, but this document may answer some incidental questions to do with parsing and rendering Pod.

In this document, "must" / "must not", "should" / "should not", and "may" have their conventional (cf. RFC 2119) meanings: "X must do Y" means that if X doesn't do Y, it's against this specification, and should really be fixed. "X should do Y" means that it's recommended, but X may fail to do Y, if there's a good reason. "X may do Y" is merely a note that X can do Y at will (although it is up to the reader to detect any connotation of "and I think it would be nice if X did Y" versus "it wouldn't really bother me if X did Y").

Notably, when I say "the parser should do Y", the parser may fail to do Y, if the calling application explicitly requests that the parser not do Y. I often phrase this as "the parser should, by default, do Y." This doesn't require the parser to provide an option for turning off whatever feature Y is (like expanding tabs in verbatim paragraphs), although it implicates that such an option may be provided.

Pod Definitions

Pod is embedded in files, typically Perl source files -- although you can write a file that's nothing but Pod.

A line in a file consists of zero or more non-newline characters, terminated by either a newline or the end of the file.

A newline sequence is usually a platform-dependent concept, but Pod parsers should understand it to mean any of CR (ASCII 13), LF (ASCII 10), or a CRLF (ASCII 13 followed immediately by ASCII 10), in addition to any other system-specific meaning. The first CR/CRLF/LF sequence in the file may be used as the basis for identifying the newline sequence for parsing the rest of the file.

A blank line is a line consisting entirely of zero or more spaces (ASCII 32) or tabs (ASCII 9), and terminated by a newline or end-of-file. A non-blank line is a line containing one or more characters other than space or tab (and terminated by a newline or end-of-file).

(Note: Many older Pod parsers did not accept a line consisting of spaces/tabs and then a newline as a blank line -- the only lines they considered blank were lines consisting of no characters at all, terminated by a newline.)

Whitespace is used in this document as a blanket term for spaces, tabs, and newline sequences. (By itself, this term usually refers to literal whitespace. That is, sequences of whitespace characters in Pod source, as opposed to "E<32>", which is a formatting code that denotes a whitespace character.)

A Pod parser is a module meant for parsing Pod (regardless of whether this involves calling callbacks or building a parse tree or directly formatting it). A Pod formatter (or Pod translator) is a module or program that converts Pod to some other format (HTML, plaintext, TeX, PostScript, RTF). A Pod processor might be a formatter or translator, or might be a program that does something else with the Pod (like counting words, scanning for index points, etc.).

Pod content is contained in Pod blocks. A Pod block starts with a line that matches <m/\A=[a-zA-Z]/>, and continues up to the next line that matches m/\A=cut/ -- or up to the end of the file, if there is no m/\A=cut/ line.

Within a Pod block, there are Pod paragraphs. A Pod paragraph consists of non-blank lines of text, separated by one or more blank lines.

For purposes of Pod processing, there are four types of paragraphs in a Pod block:

A command paragraph (also called a "directive"). The first line of this paragraph must match m/\A=[a-zA-Z]/. Command paragraphs are typically one line, as in:
```
  =head1 NOTES
```
```
  =item *
```
But they may span several (non-blank) lines:
```
  =for comment
  Hm, I wonder what it would look like if
  you tried to write a BNF for Pod from this.
```
```
  =head3 Dr. Strangelove, or: How I Learned to
  Stop Worrying and Love the Bomb
```
Some command paragraphs allow formatting codes in their content (i.e., after the part that matches m/\A=[a-zA-Z]\S*\s*/), as in:
```
  =head1 Did You Remember to C<use strict;>?
  
```
In other words, the Pod processing handler for "head1" will apply the same processing to "Did You Remember to C<use strict;>?" that it would to an ordinary paragraph -- i.e., formatting codes (like "C<...>") are parsed and presumably formatted appropriately, and whitespace in the form of literal spaces and/or tabs is not significant.
A verbatim paragraph. The first line of this paragraph must be a literal space or tab, and this paragraph must not be inside a "=begin identifier", ... "=end identifier" sequence unless "identifier" begins with a colon (":"). That is, if a paragraph starts with a literal space or tab, but is inside a "=begin identifier", ... "=end identifier" region, then it's a data paragraph, unless "identifier" begins with a colon.

Whitespace is significant in verbatim paragraphs (although, in processing, tabs are probably expanded).
An ordinary paragraph. A paragraph is an ordinary paragraph if its first line matches neither m/\A=[a-zA-Z]/ nor m/\A[ \t]/, and if it's not inside a "=begin identifier", ... "=end identifier" sequence unless "identifier" begins with a colon (":").
A data paragraph. This is a paragraph that is inside a "=begin identifier" ... "=end identifier" sequence where "identifier" does not begin with a literal colon (":"). In some sense, a data paragraph is not part of Pod at all (i.e., effectively it's "out-of-band"), since it's not subject to most kinds of Pod parsing; but it is specified here, since Pod parsers need to be able to call an event for it, or store it in some form in a parse tree, or at least just parse around it.

For example: consider the following paragraphs:

  # <- that's the 0th column

  =head1 Foo

  Stuff

    $foo->bar

  =cut

Here, "=head1 Foo" and "=cut" are command paragraphs because the first line of each matches m/\A=[a-zA-Z]/. "[space][space]$foo->bar" is a verbatim paragraph, because its first line starts with a literal whitespace character (and there's no "=begin"..."=end" region around).

The "=begin identifier" ... "=end identifier" commands stop paragraphs that they surround from being parsed as data or verbatim paragraphs, if identifier doesn't begin with a colon. This is discussed in detail in the section About Data Paragraphs and "=begin/=end" Regions.

Pod Commands

This section is intended to supplement and clarify the discussion in Command Paragraph in the perlpod manpage. These are the currently recognized Pod commands:

"=head1", "=head2", "=head3", "=head4"

This command indicates that the text in the remainder of the paragraph is a heading. That text may contain formatting codes. Examples:

  =head1 Object Attributes

  =head3 What B<Not> to Do!

"=pod"

This command indicates that this paragraph begins a Pod block. (If we are already in the middle of a Pod block, this command has no effect at all.) If there is any text in this command paragraph after "=pod", it must be ignored. Examples:

  =pod

  This is a plain Pod paragraph.

  =pod This text is ignored.

"=cut"

This command indicates that this line is the end of this previously started Pod block. If there is any text after "=cut" on the line, it must be ignored. Examples:

  =cut

  =cut The documentation ends here.

  =cut
  # This is the first line of program text.
  sub foo { # This is the second.

It is an error to try to start a Pod block with a "=cut" command. In that case, the Pod processor must halt parsing of the input file, and must by default emit a warning.

"=over"

This command indicates that this is the start of a list/indent region. If there is any text following the "=over", it must consist of only a nonzero positive numeral. The semantics of this numeral is explained in the About =over...=back Regions section, further below. Formatting codes are not expanded. Examples:

  =over 3

  =over 3.5

  =over

"=item"

This command indicates that an item in a list begins here. Formatting codes are processed. The semantics of the (optional) text in the remainder of this paragraph are explained in the About =over...=back Regions section, further below. Examples:

  =item

  =item *

  =item      *

  =item 14

  =item   3.

  =item C<< $thing->stuff(I<dodad>) >>

  =item For transporting us beyond seas to be tried for pretended
  offenses

  =item He is at this time transporting large armies of foreign
  mercenaries to complete the works of death, desolation and
  tyranny, already begun with circumstances of cruelty and perfidy
  scarcely paralleled in the most barbarous ages, and totally
  unworthy the head of a civilized nation.

"=back"

This command indicates that this is the end of the region begun by the most recent "=over" command. It permits no text after the "=back" command.

"=begin formatname"

This marks the following paragraphs (until the matching "=end formatname") as being for some special kind of processing. Unless "formatname" begins with a colon, the contained non-command paragraphs are data paragraphs. But if "formatname" does begin with a colon, then non-command paragraphs are ordinary paragraphs or data paragraphs. This is discussed in detail in the section About Data Paragraphs and "=begin/=end" Regions.

It is advised that formatnames match the regexp m/\A:?[-a-zA-Z0-9_]+\z/. Implementors should anticipate future expansion in the semantics and syntax of the first parameter to "=begin"/"=end"/"=for".

"=end formatname"

This marks the end of the region opened by the matching "=begin formatname" region. If "formatname" is not the formatname of the most recent open "=begin formatname" region, then this is an error, and must generate an error message. This is discussed in detail in the section About Data Paragraphs and "=begin/=end" Regions.

"=for formatname text..."

This is synonymous with:

     =begin formatname

     text...

     =end formatname

That is, it creates a region consisting of a single paragraph; that paragraph is to be treated as a normal paragraph if "formatname" begins with a ":"; if "formatname" doesn't begin with a colon, then "text..." will constitute a data paragraph. There is no way to use "=for formatname text..." to express "text..." as a verbatim paragraph.

"=encoding encodingname"

This command, which should occur early in the document (at least before any non-US-ASCII data!), declares that this document is encoded in the encoding encodingname, which must be an encoding name that the Encoding manpage recognizes. (Encoding's list of supported encodings, in the Encoding::Supported manpage, is useful here.) If the Pod parser cannot decode the declared encoding, it should emit a warning and may abort parsing the document altogether.

A document having more than one "=encoding" line should be considered an error. Pod processors may silently tolerate this if the not-first "=encoding" lines are just duplicates of the first one (e.g., if there's a "=use utf8" line, and later on another "=use utf8" line). But Pod processors should complain if there are contradictory "=encoding" lines in the same document (e.g., if there is a "=encoding utf8" early in the document and "=encoding big5" later). Pod processors that recognize BOMs may also complain if they see an "=encoding" line that contradicts the BOM (e.g., if a document with a UTF-16LE BOM has an "=encoding shiftjis" line).

If a Pod processor sees any command other than the ones listed above (like "=head", or "=haed1", or "=stuff", or "=cuttlefish", or "=w123"), that processor must by default treat this as an error. It must not process the paragraph beginning with that command, must by default warn of this as an error, and may abort the parse. A Pod parser may allow a way for particular applications to add to the above list of known commands, and to stipulate, for each additional command, whether formatting codes should be processed.

Future versions of this specification may add additional commands.

Pod Formatting Codes

(Note that in previous drafts of this document and of perlpod, formatting codes were referred to as "interior sequences", and this term may still be found in the documentation for Pod parsers, and in error messages from Pod processors.)

There are two syntaxes for formatting codes:

A formatting code starts with a capital letter (just US-ASCII [A-Z]) followed by a "<", any number of characters, and ending with the first matching ">". Examples:
```
    That's what I<you> think!
```
```
    What's C<dump()> for?
```
```
    X<C<chmod> and C<unlink()> Under Different Operating Systems>
```
A formatting code starts with a capital letter (just US-ASCII [A-Z]) followed by two or more "<"'s, one or more whitespace characters, any number of characters, one or more whitespace characters, and ending with the first matching sequence of two or more ">"'s, where the number of ">"'s equals the number of "<"'s in the opening of this formatting code. Examples:
```
    That's what I<< you >> think!
```
```
    C<<< open(X, ">>thing.dat") || die $! >>>
```
```
    B<< $foo->bar(); >>
```
With this syntax, the whitespace character(s) after the "C<<<" and before the ">>" (or whatever letter) are not renderable -- they do not signify whitespace, are merely part of the formatting codes themselves. That is, these are all synonymous:
```
    C<thing>
    C<< thing >>
    C<<           thing     >>
    C<<<   thing >>>
    C<<<<
    thing
               >>>>
```
and so on.

In parsing Pod, a notably tricky part is the correct parsing of (potentially nested!) formatting codes. Implementors should consult the code in the parse_text routine in Pod::Parser as an example of a correct implementation.

I<text> -- italic text

See the brief discussion in Formatting Codes in the perlpod manpage.

B<text> -- bold text

See the brief discussion in Formatting Codes in the perlpod manpage.

C<code> -- code text

See the brief discussion in Formatting Codes in the perlpod manpage.

F<filename> -- style for filenames

See the brief discussion in Formatting Codes in the perlpod manpage.

X<topic name> -- an index entry

See the brief discussion in Formatting Codes in the perlpod manpage.

This code is unusual in that most formatters completely discard this code and its content. Other formatters will render it with invisible codes that can be used in building an index of the current document.

Z<> -- a null (zero-effect) formatting code

Discussed briefly in Formatting Codes in the perlpod manpage.

This code is unusual is that it should have no content. That is, a processor may complain if it sees Z<potatoes>. Whether or not it complains, the potatoes text should ignored.

L<name> -- a hyperlink

The complicated syntaxes of this code are discussed at length in Formatting Codes in the perlpod manpage, and implementation details are discussed below, in About L<...> Codes. Parsing the contents of L<content> is tricky. Notably, the content has to be checked for whether it looks like a URL, or whether it has to be split on literal "|" and/or "/" (in the right order!), and so on, before E<...> codes are resolved.

E<escape> -- a character escape

See Formatting Codes in the perlpod manpage, and several points in Notes on Implementing Pod Processors.

S<text> -- text contains non-breaking spaces

This formatting code is syntactically simple, but semantically complex. What it means is that each space in the printable content of this code signifies a non-breaking space.

Consider:

    C<$x ? $y    :  $z>

    S<C<$x ? $y     :  $z>>

Both signify the monospace (c[ode] style) text consisting of "$x", one space, "?", one space, ":", one space, "$z". The difference is that in the latter, with the S code, those spaces are not "normal" spaces, but instead are non-breaking spaces.

If a Pod processor sees any formatting code other than the ones listed above (as in "N<...>", or "Q<...>", etc.), that processor must by default treat this as an error. A Pod parser may allow a way for particular applications to add to the above list of known formatting codes; a Pod parser might even allow a way to stipulate, for each additional command, whether it requires some form of special processing, as L<...> does.

Future versions of this specification may add additional formatting codes.

Historical note: A few older Pod processors would not see a ">" as closing a "C<" code, if the ">" was immediately preceded by a "-". This was so that this:

    C<$foo->bar>

would parse as equivalent to this:

    C<$foo-E<gt>bar>

instead of as equivalent to a "C" formatting code containing only "$foo-", and then a "bar>" outside the "C" formatting code. This problem has since been solved by the addition of syntaxes like this:

    C<< $foo->bar >>

Compliant parsers must not treat "->" as special.

Formatting codes absolutely cannot span paragraphs. If a code is opened in one paragraph, and no closing code is found by the end of that paragraph, the Pod parser must close that formatting code, and should complain (as in "Unterminated I code in the paragraph starting at line 123: 'Time objects are not...'"). So these two paragraphs:

  I<I told you not to do this!

  Don't make me say it again!>

...must not be parsed as two paragraphs in italics (with the I code starting in one paragraph and starting in another.) Instead, the first paragraph should generate a warning, but that aside, the above code must parse as if it were:

  I<I told you not to do this!>

  Don't make me say it again!E<gt>

(In SGMLish jargon, all Pod commands are like block-level elements, whereas all Pod formatting codes are like inline-level elements.)

Notes on Implementing Pod Processors

The following is a long section of miscellaneous requirements and suggestions to do with Pod processing.

Pod formatters should tolerate lines in verbatim blocks that are of any length, even if that means having to break them (possibly several times, for very long lines) to avoid text running off the side of the page. Pod formatters may warn of such line-breaking. Such warnings are particularly appropriate for lines are over 100 characters long, which are usually not intentional.
Pod parsers must recognize all of the three well-known newline formats: CR, LF, and CRLF. See perlport.
Pod parsers should accept input lines that are of any length.
Since Perl recognizes a Unicode Byte Order Mark at the start of files as signaling that the file is Unicode encoded as in UTF-16 (whether big-endian or little-endian) or UTF-8, Pod parsers should do the same. Otherwise, the character encoding should be understood as being UTF-8 if the first highbit byte sequence in the file seems valid as a UTF-8 sequence, or otherwise as Latin-1.

Future versions of this specification may specify how Pod can accept other encodings. Presumably treatment of other encodings in Pod parsing would be as in XML parsing: whatever the encoding declared by a particular Pod file, content is to be stored in memory as Unicode characters.
The well known Unicode Byte Order Marks are as follows: if the file begins with the two literal byte values 0xFE 0xFF, this is the BOM for big-endian UTF-16. If the file begins with the two literal byte value 0xFF 0xFE, this is the BOM for little-endian UTF-16. If the file begins with the three literal byte values 0xEF 0xBB 0xBF, this is the BOM for UTF-8.
A naive but sufficient heuristic for testing the first highbit byte-sequence in a BOM-less file (whether in code or in Pod!), to see whether that sequence is valid as UTF-8 (RFC 2279) is to check whether that the first byte in the sequence is in the range 0xC0 - 0xFD and whether the next byte is in the range 0x80 - 0xBF. If so, the parser may conclude that this file is in UTF-8, and all highbit sequences in the file should be assumed to be UTF-8. Otherwise the parser should treat the file as being in Latin-1. In the unlikely circumstance that the first highbit sequence in a truly non-UTF-8 file happens to appear to be UTF-8, one can cater to our heuristic (as well as any more intelligent heuristic) by prefacing that line with a comment line containing a highbit sequence that is clearly not valid as UTF-8. A line consisting of simply "#", an e-acute, and any non-highbit byte, is sufficient to establish this file's encoding.
This document's r