Humdrum encodings consist of a set of one or more lines or records. There are three types of Humdrum records:

• comment records,
• interpretation records, and
• data records.

These three record types are mutually exclusive, so it is not possible to mix comments, interpretations, or data records on the same line.

Comment Records

Interpretation Records

Interpretations are lines that begin with the asterisk character (*). Interpretations are used to identify more precisely the state of the representation -- for example, to indicate that an encoded part is for a transposing instrument in E-flat, or to indicate that the representation is for a given Balinese tuning, or that the representation encodes a conductor's physical gestures. Humdrum requires that at least one interpretation must be specified before any data records are encountered. The difference between a comment and an interpretation is that interpretations are formal, potentially executable statements; interpretations pass information to programs that process the Humdrum encoding.

As in the case of comments, there are two types of interpretations: exclusive interpretations begin with a double asterisk (**) whereas tandem interpretations begin with a single asterisk (*). Exclusive interpretations are mutually exclusive: only one such interpretation can be active at a given time for a given string of data. No set of data is complete without the presence of an exclusive interpretation. Tandem interpretations, by contrast, provide supplementary information about how a set of data is to be interpreted. Several tandem interpretations may pertain to a given set of data; unlike exclusive interpretations, tandem interpretations are not necessarily mutually exclusive.

Data Records

Lines that do not contain either an exclamation mark or an asterisk in the first column are data records. Blank lines (i.e., lines which are either empty, or contain only blank space, such as tabs and spaces) are forbidden in Humdrum. Thus data records may be formally defined as non-empty lines that do not begin with either an exclamation mark or an asterisk.

In Humdrum, each data record encodes information pertaining either to a particular moment in time or to a particular time window or duration. (Whether a record represents a precise moment or an expanse of time depends on the accompanying interpretation.) Each data record may contain one or more data tokens. When more than one token is present, tokens are separated from each other by tabs. When several data records are present, multiple tokens are aligned in columns through the file. As we noted earlier, these columns are referred to as spines. By itself, a spine has no particular meaning; it is simply a way of linking together related tokens through time. Spines become meaningful only when they are labelled by adding an interpretation.

By itself, Humdrum recognizes only six ASCII characters. Two of these characters -- the exclamation mark (!) and the asterisk (*) -- have a special meaning only when they appear in the first column of a record (or are preceded by a tab; see below). The remaining special Humdrum characters are the period (.), the space, the tab character, and the carriage return (= newline character). As we have seen, the exclamation mark and the asterisk are used to identify comments and interpretations, respectively. The tab and carriage return characters are used to format the data into spines and records, respectively.

Data Tokens and Null Tokens

As we noted above, the data in the data records are conceptually divided into tokens. In Humdrum, there are two possible types of tokens:

• data tokens, and
• null tokens (.).

Interpretations can be cascaded so that a single spine has more than one interpretation associated with it. This is done through the addition of tandem interpretations. Consider the following example:

**foot	**foot	**arm	**arm
*left	*right	*left	*right
X	.	.	X
.	X	X	.
X	.	.	X
.	X	X	.
X	.	.	X
*-	*-	*-	*-

Users are free to define as many different exclusive and tandem interpretations as they wish. For example, a user might define the interpretation **bowing that would be suitable for encoding detailed bowing information in works for strings. For each exclusive interpretation, the Humdrum user can re-define the meaning of all of the text characters, with the exception of the tab and the carriage return, which always retain their functions as `token/spine separator' and `record separator' respectively. The characters ! . * can also be re-defined, although there are some restrictions as to how they can be used. Specifically, the exclamation mark cannot occur in the first column of the record unless it is used to indicate a comment. Similarly, the asterisk cannot occur in the first column of a record unless it is used to indicate a Humdrum interpretation. The period cannot appear in the first column unless it is used to indicate a null data token. In addition, the exclamation mark, asterisk, and period cannot appear following a tab unless they are used to indicate a comment, interpretation, or null token, respectively.

Data Sub-Tokens

Data tokens can be split into sub-tokens via the space character. In the first data record of the following example, the first spine contains two sub-tokens whereas the third spine contains three sub-tokens. Sub-tokens do not have their own spine organization and can appear and disappear as necessary:

**spine1	**spine2	**spine3
A B	J	X Y Z
AB	J	XYZ
A B C	.	X Z
*-	*-	*-

Data sub-tokens are useful in a variety of circumstances. An appropriate use of sub-tokens might be to encode double- and triple-stops in string parts.

In the Humdrum data records, the space character is reserved solely for use as a sub-token delimiter. Note that consecutive spaces are illegal, and that data tokens cannot begin or end with a space character. Of course spaces can be used freely in comments and in interpretations.

Spine Paths

Humdrum representations often consist of a fixed number of spines that continue throughout the course of an encoded file. As we have seen in the preceding chapters, a typical use of spines is to encode different voices or parts in a musical work. However, there is no reason to equate spines with voices; spines are used for many other purposes as well.

In encoding Humdrum representations it is occasionally useful to be able to vary the number of spines. However, files with varying numbers of spines can pose significant questions of interpretation. Consider, for example, the following sequence of Humdrum-like data records:

1			2			3
1			2			3
1			2			3
A			B
A			B
A			B

The Humdrum syntax provides special spine path indicators that make it possible to resolve such ambiguities and to ensure that the continuity (or lack of continuity) is made clear. Humdrum provides five special path indicators, one of which we have already encountered:

• a new spine may be introduced
• an existing spine may terminate (without continuing further)
• a previous spine may be split into two spines
• two or more spines may be amalgamated into a single spine
• the positions of two spines may be exchanged

*+	add a new spine (to the right of the current spine)
*-	terminate a current spine
*^	split a spine (into two)
*v	join (two or more) spines into one
*x	exchange the position of two spines
*	null interpretation (place holder)

1	2	3
*	*-	*		(elimination of spine #2)
1	3
1	2	3
*	*x	*x		(exchange spines #2 and #3)
1	3	2
1	2	3
*	*^	*		(splitting of spine #2)
1	2a	2b	3
1	2	3
*	*v	*v		(amalgamation of spines #2 and #3)
1	2&3

1	2	3
*	*+	*		(add a new spine.)
*	*	**inter	*	(define exclusive interpretation for the new spine)
1	2	new	3

The following examples illustrate a variety of more complex path redefinitions:

1	2	3	4
*v	*v	*^	*^
1&2	3a	3b	4a	4b

1	2	3	4	5
*	*-	*	*-	*
*v	*v	*v
1&3&5

1	2	3	4	5
*	*-	*	*^	*+
*	*	*	*	*	**new
*v	*v	*	*	*	*
1&3	4a	4b	5	new

1	2	3	4
*x	*x	*	*
*	*x	*x	*
*	*	*x	*x
2	3	4	1

Syntactically, some path constructions are illegal; here are some examples of illegal constructions:

1	2	3
*v	*	*v		(The join-spine indication in spine #1 does not adjoin
				spine #3.)
1	2	3
*x	*x	*x		(No more than two exchange interpretations at a time.)
1	2	3
*x	*	*		(Must have two exchange interpretations together.)
1	2	3
*v	*	*		(Must have two or more join interpretations at a time.)
1	2	3
*	*			(Spine eliminated without using a termination interpretation.)
1	2
1	2	3
*	*	*+		(Adding a new spine should result in 4 interpretations.)
1	2	3
1	2
*	*	*-		(Cannot eliminate non-existent spine.)
1	2
*+	*
1	new	2		(New spine started without specifying new interpretation.)
1	2
*	*+
*	**inter	*		(Interpretation labels the wrong spine.)
A	B	C

With the preceding background it is now possible to define formally a Humdrum representation. First we can define a Humdrum file. A Humdrum file must conform to one of the following:

• A file containing comments, data records and interpretations with the restriction that no data record or local comment appears before the first exclusive interpretation.
• A file containing data records preceded by at least one exclusive interpretation.
• A file containing only comments and interpretations with the restriction that no local comments appear before the first interpretation.
• A file containing only interpretations beginning with an exclusive interpretation.
• A file containing only global comments.
• A totally empty file (i.e. a file containing no records).

Additional interpretations may be added throughout the file. Global comments may appear anywhere in the file. However, local comments are much more restricted: (1) Local comments may not appear until after the first interpretation record, and (2) The number of sub-comments in a local comment record must be equivalent to the number of currently active spines.

Comment	Either a global or local comment. Any record beginning
	with an exclamation mark.
Global comment	Any record beginning with two exclamation marks (!!).
Local comment	Any record beginning with one and only one exclamation mark (!).
	Every spine in that record must also begin with an exclamation
	mark.
Null comment	A comment record containing no commentary; only the
	appropriate exclamation mark(s) are present.
Interpretation	Either an exclusive or tandem interpretation. Any record
	beginning with an asterisk (*).
Exclusive interpretation	Any record beginning with one or more asterisks (*), where at
	least one spine begins with two asterisks.
Tandem interpretation	Any record beginning with a single asterisk (*) where none of the
	spines begin with two asterisks.
Path indicator	One of five special tandem interpretations *+ *- *v *^ *x found only
	in tandem interpretation records.
Null interpretation	An interpretation for a given spine or spines consisting of just the
	interpretation signifier (i.e., a single asterisk).
Data record	Any record that is not a comment or interpretation. Must contain
	the same number of tokens as the number of current spines.
Null token	The period (.) either alone on a single record or separated from
	other characters by a tab. Appears only in data records.
Null data record	A data record consisting only of null tokens.
Spine	A column-like "path" of information -- including data records,
	local comments, and interpretations.

An empty record.

A record containing only tabs.

A record beginning with a tab.

A record ending with a tab.

Any record containing two successive tab characters.

Any data record having fewer or more spines than the immediately

preceding data record.

A record having only one join-spine indication.

A record having only one exchange-spine indication.

A record having more than two exchange-spine indications.

Some Illegal Humdrum Constructions

The humdrum Command

Reprise

This chapter has identified the formal structural and organizational features of the Humdrum syntax. The syntax provides a framework within which sequential symbolic data can be represented. Individual representation schemes map the ASCII character set (signifiers) to various music-related concepts (signifieds).

Each representation is designated by an exclusive interpretation. The corresponding data are organized in spines that may meander throughout the file. New spines may be added, spines joined together, exchanged, split, or terminated. Data are organized as tokens, although tokens can consist of multiple subtokens separated by single spaces. Null tokens can appear as place-holders where no specific data exists.

Free-form comments may be interspersed throughout the file. Global comments pertain to all spines whereas local comments pertain to individual spines. Additional interpretive information may be encoded using tandem interpretations. Both local comments and tandem interpretations may occur anywhere, but must be preceded in the spine by some exclusive interpretation.