Re: First/Follow sets in Recursive Descent Parsers

The simplest way is to replace the
    integral = '\int' [subscript superscript] '{' expression 'd' VARIABLE
'}'.


with


integral = '\int' [subscript superscript] '{' expression VARIABLE VARIABLE
'}'.


and test the first variable for 'd' in the code, giving a semantic error if
it isn't.


Tom




"Dan Haxell" <Dan.Haxell@btinternet.com> wrote
>I am currently writing a utility to translate LaTeX math expressions
>into Maple syntax. So far I have been using Flex to generate the
>lexical analyser and then hand-coding a recursive descent parser (in
>C++).
>
>It is based on the following grammar expressed in EBNF:
>
>expression = ['+' | '-'] term { ('+' | '-') term }.
>term = factor ( { ('*' | '/') factor } | { factor } ).
>factor = entity [ '^' '{' expression '}' ].
>entity = NUMBER | VARIABLE [ '(' expression ')' ] | '(' expression ')' |
>integral.
>integral = '\int' [subscript superscript] '{' expression 'd' VARIABLE '}'.
>subscript = '_' ( '{' expression '}' | (VARIABLE | NUMBER) ).
>superscript = '^' ( '{' expression '}' | (VARIABLE | NUMBER) ).
>
>Using this, strings of letters such as 'abc' are translated as products:
>'a*b*c'.
>
>Each non-terminal is implemented as a void function that returns a
>string containing the Maple translation, e.g. string factor(void).
>
>The problem I have is when parsing an integral: during the parse, I
>want the letter 'd' to have a new meaning, namely to designate the end
>of the expression and indicate the variable to integrate. This should
>only be the case when parsing an integral and normally it should be
>treated as a simple variable.
>
>However, when parsing the expression 2x dx, the 'dx' is not recognised
>as being special since the program is still in the Expression()
>function and thus parses as 2*x*d*x.
>
>Would passing First/Follow sets (stop sets??) to each function allow
>for situations such as this? Many of the examples of RDPs use void
>functions as well.

Return to the comp.compilers page.
Search the comp.compilers archives again.