readme

现将本人搜集常用的程序设计语言的词法和形式文法提供给大家。其中

This is a Fortran 77 subset parser that I knocked together in about a
week.  It does a decent job of tokenizing and parsing Fortran,
although it is quite incomplete.  It is provided as is without any
claim that it is suitable for any purpose nor with any promise of
support.  Copying is permitted so long as it is not for direct
commercial advantage.  See the copyright notice in each source file.

The general strategy is as follows:

	1.  Read each line, determine if it's a comment or
	continutation.
	2.  If not a comment, add into the current statement buffer,
	removing quoted literals and hollerith strings and replacing
	them by special tokens (a space followed by a digit.)
	3.  Once we have an entire statement, scan it to see if it
	looks like an assignment statement.  If not it must be a
	keyword	statement.
	4.  Call the yacc parser.  From this point, tokenizing is
	pretty simple -- see if any of the keyword or special
	character tokens that are allowed in the current context
	match, if not look for a number or a name. Statement numbers
	are treated as integer constants.

The one case where I know that I parse wrong is something like:

	real function a(34)

which dimensions the array functiona but parses incorrectly as a function
declaration with a bogus dummy argument.  This is hard to fix and is unlikely
to cause trouble in practice.

The language it parses includes some invalid constructs that would
need to be kicked out semantically, e.g.:

	dimension foo

because I treat dimension and explicit type statements the same.

If you want to extend this to parse more of Fortran, you'll obviously
need to add more yacc syntax (note that there is syntax in the
current parser that the lexer doesn't generate the tokens for,) more
lexer tokens, and more lexer states.  Lexer states you'll need
certainly include:

	-- Looking for the TO in an assign statement.
	-- Tokenizing format statements (unless you want to do
	it at runtime, a common approach)
	-- Looking for the zillion possible keywords in I/O
	statements.

There are doubtless more, you'll know them when you see them.

I have run this under Microsoft C 5.1 on a PC and under GCC 1.35.  It
should work on the regular unix PCC with minimal modifications,
mostly getting rid of a few ANSI function prototypes that I use and
perhaps adjusting the varargs constructs in emit().  It has parsed
many of the old IBM SSP routines, so I'm fairly confident that it
parses what it parses pretty well.

If you do anything interesting with this, I'd appreciate hearing
about it.

John Levine
P.O. Box 349
Cambridge MA 02238
[phone]
Internet: [e-mail] or [e-mail]
MCI Mail: 103-7498, WUI telex 6501037498 MCI UW
Genie: J.LEVINE3