Appendix A2.1.1. Backus–Naur form of the STAR syntax and grammar

We accept a space, a horizontal and a vertical tab as 〈blank〉.

The non-printing single ASCII characters 10, 12, 13 or any sequence of these are always interpreted as being a single line terminator. In this way there should not be operating-system-dependent ambiguity in those architectures that use character sequences as line terminators. This necessarily requires that these characters can only be used for line termination.

We define a `comment' to be initiated with 〈blank〉 or 〈terminate〉 and the character #, followed by any sequence of characters (which include 〈blank〉). The only characters not allowed are those in the production 〈terminate〉, and hence these characters terminate a comment. Note the requirement of a leading 〈blank〉 or 〈terminate〉 is dropped if the # character is the first character in the file.

We accept as white space all elements in the above three productions. White spaces are the lexemes able to delimit the lexical tokens. Note that a comment is a legitimate white space because it must end with a line terminator, and hence delimits tokens.

Non-blank characters are composed of all the characters in our set, excluding 〈blank〉 and 〈terminate〉 characters.

〈char〉 characters are composed of all the characters in our set, excluding 〈terminate〉 characters.

We define a `line of text' to be a line contained within a semicolon-bounded text block. Hence the first character cannot be a semicolon, and is followed by any number of characters from the set 〈char〉 and terminated with a line-termination character or just the termination character. This allows for `blank' lines in the semicolon-bounded text block.

Productions for specific characters.

All printable characters except the double quote.

All printable characters except the single quote.

All printable characters except the left and right square brackets.

All printable characters except the semicolon.

Ordinary characters are all those printable characters that can initiate a non-quoted text string. These exclude the special characters, ", #, $, ' and _ and in some cases ;.

The keywords (in a case-insensitive form).

The operating-system-dependent end-of-file marker.

A STAR File may be an empty file, or it may contain one or more data blocks or global blocks.

There can be any amount of white spaces (remember 〈wspace〉 includes comments) before and at least one white space or an end of file (EOF) after a data or global block. This forces white space between data (and global) blocks in a single file. There must be at least one data item in any data or global block. This means a file consisting of just a data or global block heading is invalid.

There can be any amount of white spaces (remember 〈wspace〉 includes comments) before a save-frame block. This forces white space between save-frame blocks also. There is no need to include the { 〈wspace〉+ | 〈EOF〉 } found in data and global blocks, since those productions cover the situation of a save-frame block terminating the file.

A data-block or save-frame heading consists of the relevant five-character keyword (case-insensitive) immediately followed by at least one non-blank character. This does not preclude the associated block name or frame name consisting of just one or more punctuation characters.

Data come in the following three forms.

We must allow for white space preceding the loop_ (case-insensitive) keyword, since this is not covered by any of the other productions.

The name list for a loop must include at least one data name or a nested loop.

A data name is initiated by an underscore character and followed by one or more non-blank and non-terminating characters from the STAR character set. This does not preclude data names consisting of just one or more punctuation characters.

Loop values are represented in the same way as the 〈data〉 production, except that the possibility of nested data loops introduces the need for the stop_ keyword.

Data values of type I data are immediately preceded by a 〈blank〉. Data values of type II data are immediately preceded by a 〈terminate〉.

A type-I unquoted string is immediately preceded by a 〈blank〉. It cannot begin with a number of characters (the complement of the 〈ordinary_char〉 set) i.e. ", #, $, ', [, ] and _. However, it can begin with a semicolon. Then it is followed by any number of non-blank characters.

A type-II unquoted string is immediately preceded by a line break. As with type I, it too cannot begin with a ", #, $, ', [, ] or _. It also cannot begin with a semicolon, since this would match the semicolon-delimited data production.

Specific exceptions to lexemes which match both types of unquoted strings are:

If one wishes to define data values which match lexemes excluded in cases (1) and (2) above, they should be quoted data values.

The string between a set of double quotes can consist of any character that is not a double quote, or it can be a double quote as long as it is immediately followed by a non-blank character or any number of double quotes at the end of the string. This final rule picks up cases of double-quote delimited strings that end in one or more double quotes, like "ABC"".

The string between a set of single quotes can consist of any character that is not a single quote, or it can be a single quote as long as it is immediately followed by a non-blank character or any number of single quotes at the end of the string. This final rule picks up cases of single-quote delimited strings that end in one or more single quotes, like 'ABC''.

The string bounded by semicolons can begin with any number of characters (including those in the 〈blank〉 production) but is necessarily terminated by a line break. This forces a line break on the line that contains the `opening' semicolon. After the first line, one can have any number of 〈line_of_text〉. Note we treat the first line as special, since it can contain a leading semicolon, which is not true of 〈line_of_text〉. A 〈line_of_text〉 is always terminated with a line break, thus ensuring the closing semicolon is in column 1.

The string bounded by square brackets can consist of any character including 〈terminate〉 and 〈blank〉, and excluding the characters [ and ] unless they are escaped or are balanced.