In the previous phase of the class project, you used the flex tool to create a lexical analyzer for the "MINI-L" programming language that reads in a MINI-L program (from a text file) and identifies the sequence of lexical tokens in the text file. In this phase of the class project, you will create a parser using the bison tool that will check to see whether the identified sequence of tokens adheres to the specified grammar of MINI-L. The output of your parser will be the list of productions used during the parsing process. If any syntax errors are encountered during parsing, your parser should emit appropriate error messages. Additionally, you will be required to submit the grammar for MINI-L that you will need to write before you can use bison.
[The MINI-L language is described in detail here. Pay particular attention to the
syntax diagrams here that describe the appropriate structure of the various MINI-L statements.]
[The required output format for your parser is described here.]
Bison is a tool for generating parsers. Given a specified context-free grammar for a language, bison will automatically produce an LALR(1) bottom-up parser for that language. The parser will ensure that a given sequence of tokens are ordered in such a way that they adhere to the specified grammar of the language. If the tokens fail to parse properly, then appropriate syntax error messages are outputted.
Shift/Reduce and Reduce/Reduce conflicts:
When ambiguities exist in the specified grammar, bison will emit one or more
conflicts when it is run. These conflicts do not prevent bison from
generating the parser from your specification, but they may unexpectedly affect how your parser
behaves. A shift/reduce conflict occurs when the parser may
perform either a shift or a reduce. A reduce/reduce conflict occurs when
there are two or more production rules that apply to the same sequence of input tokens.
In general, reduce/reduce conflicts indicate errors in the grammar (or at least serious
ambiguities) and should be eliminated by modifying the grammar specification as needed.
Shift/reduce conflicts, on the other hand, are more difficult to completely eliminate,
especially when using the special "error" token provided by bison to handle
errors. Therefore, you should try to eliminate as many shift/reduce conflicts as you
can, but some shift/reduce conflicts may remain as long as they do not adversely affect
the behavior of your parser. You can run bison with
the -v option to generate an output file containing detailed information about
each conflict.
In our department, bison is installed and can be used on machine "bass".
[A brief introduction to bison can be found here.]
[The detailed manual for bison can be found here.]
mini_l.y.
-d flag is necessary to create a .h file that will link
your flex lexical analyzer and your bison parser. The -v
flag is helpful for creating an output file that can be used to analyze any conflicts
in bison. The --file-prefix option can be used to
change the prefix of the file names outputted by bison. bison -v -d --file-prefix=y mini_l.y. This will
create the parser in a file called y.tab.c, the necessary .h file
called y.tab.h, and the informative output file called y.output.
mini_l.lex, then
use it with flex as follows: flex mini_l.lex. This will create the
lexical analyzer in a file called lex.yy.c.
parser
with the following command: gcc -o parser y.tab.c lex.yy.c -lfl.
The program parser should now be available for use.
Suppose your parser is in the executable named parser. Then for
the MINI-L program primes.min, your parser should be
invoked as follows:
cat primes.min | parser
The list of productions taken by your parser during parsing should then be printed to the screen. As one example, the output might look like this (you do not need to number each production or label each non-terminal with the corresponding production number). However, your output may be different due to different productions in your specification. The most important thing is that your parser should not output any syntax errors for syntactically correct programs, and your parser should output helpful syntax error messages (for at least the first syntax error) whenever syntax errors do exist in the inputted MINI-L program.
Submit via Moodle. Turn in your source code and the Makefile (or a build script) in a .tar.gz or .zip file. Make sure you can uncompress and compile your submission on bass, otherwise your score for this project will be 0.
Similar to project 1, if you implement this project in Objective Caml (an ML variant that supports functional, object-oriented, and imperative programming) you will receive 10% extra credit on this project's score.
We have installed OCaml on bass. It comes with its own parser generator, ocamlyacc (similar to bison), and an impressive array of libraries.
If you are having technical difficulties with OCaml, contact the instructor. TA's do not have time and resources to help students with the OCaml implementation.