* Lars request (see dir contributions): allow AST classes to be Moose or any other OOP framework
Node clases inherit from Parse::Eyapp::Node.pm inherit from Moose, wrapping around Node?
* Lexer generation:
%token = /(reg1)(reg)(reg3)/
return the array of parenthesis not just $1
* Include PERLCODE as a predefined token in lexer generation
* La informacion acerca de los números de línea en el debugger
para la generación del analizador léxico esta hecha un desastre
* Bug and conceptual errors in contextual tokens: (see t/77. I believe it is solved)
* factorize, factorize makeLexer!!!!
* There is a "token" synchronization problem in this case
See files:
t/forforeacherikcontextual.eyp
t/forforeacherik.eyp
t/forforeacherikcontextual2.eyp
t/C.eyp
creo que hay salvar en self->{POS} la $pos despues del matching
en el código del lexer generado para este tipo de tokens
* %/.../ defines contextual tokens and it is independent of
/ ... /=Var
/ ... /!Var
they must exist independently
* This technique can be used to parse difficult ambiguous grammars
EQUAB = %//=ab
then in the body of the grammar:
s: EQUAB ab cs | as bc
etc.
* Check that Driver and generated parser are compatible. Version numbers
* Implement support for tie-ins in the generated lexical analyzer
* AmbiguousLanguage2.eyp usar como contraejemplo a la primera teoria
The difficulty of using yacc is extraordinary. See t/noPackratSolvedExpRG2.yp
x:
NUM
| NUM OP x
making it left recursive ruines the whole plan
* Do not introduce productions for exploring, use "explorer states" instead
* The explorer variable has to have the same name of the conflict:
sub YYNestedParse {
my $self = shift;
my $parser = shift;
my $conflictName = $self->YYLhs;
...
* Modifiy Parse::Eyapp::Driver::_Parse
If the state is a conflict state execute the conflict handlers for that state in
the presence of a conflictive token!
* Option -v in eyapp will produce a navigable grammar.output.html version of the output
file
* EBNF ors must be implemented
* Artículos:
* OOPLAS
* reescribir con nueva sintáxis TOPLAS y el comsys
* Introducir modificaciones de los referees en los articulos de revista: implementacion, pruebas
* conseguir articulo de Ganapanathi "semantic predicates in"
* estudiar e introducir comparaciones
* insistir en que esta implementado y "verificado": describir implementación
* Referees PLDI
-- too vague description
-- motivation for the YYExpected: explain the problem
-- Syntactic analysis that depends on the semantic phase: explain better
-- how the PPCR is better than backtracking
-- more mathematics, proofs,
-- One pass: is not clear to the referee
-- Doubts about the implementation: have been implemented paragraph, see example files in
such, and such addresses
-- Comparisons with other parsers: Regexp::Grammars, Marpa: execution times, etc.
Complexity.
-- typos, wording
-- two different articles
* optimize generated lexer for contextual tokens (Expected must be called only once)
* debugging: break points, call actual debugger inside semantic actions, etc.
* Trying to supress the generation of conflict rules when there is no conflict (during nested
auxiliary compilation). Count the number of appearances of %PREC conflict . If it is greater
than two there is a true conflict. Otherwise the conflict symbols can be ignored.
There is however a problem. The translation used by Parse.yp performs some premature
semantic actions that insert the conflict symbols as soon as a production is defined.
To solve it we must delay the translation until the whole grammar has been seen.
* Write code to control the right use of %conflict and %PREC
- x $self->{GRAMMAR}{CONFLICTHANDLERS}
0 HASH(0x1009898a0)
'decORexp' => HASH(0x100c5c1e0)
'explorerline' => 14
'line' => 14
'production' => HASH(0x100c50130)
'-12' => ARRAY(0x100c572e0)
0 1
'-5' => ARRAY(0x100c50230)
0 1
'states' => ARRAY(0x100c4bb50)
0 HASH(0x100c5bba0)
28 => ARRAY(0x100c3c5b0)
0 '\')\''
'total' => 2
'totalviewpoint' => 1
DB<8> x $$self{CONFLICTS}{FORCED}
0 HASH(0x100c38800)
'DETAIL' => HASH(0x100c5b630)
28 => HASH(0x100c58030)
'LIST' => ARRAY(0x100c581c0)
0 ARRAY(0x100c581e0)
0 '\')\''
1 '-12'
'TOTAL' => ARRAY(0x100c57d10)
0 empty slot
1 1
'TOTAL' => ARRAY(0x100c57da0)
0 0
1 1
- %PREC in no conflict production or in wrong position
- called for reduce-reduce but it is not a reduce-reduce, etc.
- Every %conflict has to have at least two associated %PRECs
* quiza entre la información guardada en CONFLICTHANDLERS deberían ir las etiquetas simbólicas.
Esto simplificaría YYSetReduce
* Otra API de mas alto nivel sería el caso regexp
%conflict conflictname /regexp/ ruleName
esta se ejecuta en el punto del conflicto
* clean tests deleted. Do it gain?
* examples in MANIFEST: obsolete
* Use %{ ... %} as syntactic sugar for syntactic code actions
* Generation of Several Parsers
- The lexer must be the one of the main parser
* The accept state must change
A$ -> A
A$ -> A.b where b is in the follow of A, means acceptation
Where are the follows?. They have been computed somewhere
* %PREC conflict does not check that a conflict handler has been previously declared
* GLR in eyapp
* eyapp and PPCR in the Wikipedia
* Update documentation
* add eyapp -Y option: equivalent to "%tree bypass"
* Lalr
# substitute End Of Input by DEFAULT for each state
# Will it give a false acceptance?
# I think I shall use an option in eyapp?
# make a test
* Change the API for YYSetReduce and YYSetshift to include the name of the conflict?
* Bug/limitation: When using -m and -o the output file still uses the original grammar name.
It seems more convenient to rename it to the output file suffixed with .output
* Lexical Analyzer generation:
* incremental lexer also influences the way the directive
%lexer
must behave. Because a lexer written by hand can be incremental
* support streams: see lexer in examples/Eyapp/Lhs1.yp
- Parse::Eyapp::Driver::input called as a setter receives an open file handler :
prompt!
- if the input string is finished it reads from the file handler (controled by $/)
otherwise returns the current string
* error line numbers-> source code
* lexer must support a sub name
%lexer mylexer
* Document:
=head3 $parser->input YYInput
=head3 $parser->YYInputFile
=head3 $parser->YYPrompt
=head3 $parser->main
=head3 $parser->Run
* %tree:
- info clause (autobuild ::info for TERMINAL)
- naming scheme clause
* Tests
- tests all examples/*/*yp for syntax checking
- test for yyfilename
- Add tests to check .output tables: actions and gotos against yacc.
I am not confident that some modifications I introduced in Grammar.pm were error free
- Automatic naming scheme: tests!
- Some rules can be optimized away. Then the name dissapears!!
- tests with syntax errors
- Debug mode and standalone works?
* Debugging: set breakpoints in the code from yydebug=> 1
* What if the input comes from a stream? then $parser->input is undef
Shall I support interactive streams?
- YYLookaheads assumes that input comes from $parser->input
- The default lexer assumes that input comes from $parser->input
Change it to use an input stack
* PPCR:
- The whole implementation approach seems wrong.
Make actions an array. Associate the handler with the state.
That allows a syntax like:
%conflict handler {... }
%shift-reduce prod1 [token1, token2, ...] prod2 [tokens] handler
%reduce-reduce prod3 [token1, token2, ...] prod4 [tokens] otherhandler
....
- There is a bug: I have to check that reductions in PPCR only occur inside the conflictive states
- Warning Grammar.pm
sub Parse::Eyapp::Grammar::conflictHandlers rewrite! or use Data::Dumper forever?
* %package directive: equivalent to eyapp -m option
* -%modulino directive
%modulino perlinterpreter prompt
Equivalent to eyapp -b 'perlinterpreter' and __PACKAGE__->main(prompt)
%modulino '/usr/local/lib/perl/5.10.1/bin/perl5.10.1' 'Try input "type e = (x);": '
avoid repetitive __PACKAGE->main when using options -C and -b
* Article: rewrite, both research and divulgation
* Documentation
- Simpler intro
- Separate parsing from tree transformation
* EBNF (a|b|c)*
- eyapp option -TREE: same as %tree
currently only give_rhs is used, but it has to be parametrized
* - eyapp option namingscheme
- eyapp option -start (defines the start symbol)
Eyapp -> Lalr -> Grammar -> Options
`----> Parse
- intermediate actions in give_rhs_name? check, test
* Error message when a conflict handler isn't used is inaccurate:
~/LEyapp/examples/debuggingtut$ eyapp dynamicgrammar.eyp
Useless non-terminals:
rangeORenum, declared line 10
Useless rules:
rangeORenum ->
1 shift/reduce conflict and 0 reduce/reduce conflicts
* Line numbers seem to be in a buggy state.
There is also a "bug" with tokenline and the static attribute approach: tokenline
attribute is defined to 1 while the static/attribute approach searches while "undef"
* Lars request (see dir contributions): allow AST classes to be Moose or any other OOP framework
* Provide a module providing regexp for most common non destructive lexing: NUMs, IDs, etc
use Parse::Eyapp::LexSupport qw{INT PERLWS ID};
Created by "new" so that $self is in scope
$INT = qr{\G(\d+)}gc; $WS = qr{{\G(\s*) (?{ $self->tokenline($1 =~ tr{\n}{}) }) }gxc; etc.
tokens can be "defined" with:
%token DOTDOT /(\.\.)/
%token INT /(\d+)/
The parser will define $DOTDOT = qr{/\G(\.\.)/;
and then:
PERLWS(); # No need to update the line number
m{\Gtype\b}gic and return ('TYPE', 'TYPE'); # old way
m{$ID}gc and return ('ID', $1);
INT() and return ('NUM', $1);
DOTDOT() and return ('DOTDOT', $DOTDOT);
m{\G(.)}gc and return ($1, $1);
* Parse::Eyapp::Driver::main
- option to send output to a file
* bison directives:
Directive: %debug
In the parser file, define the macro YYDEBUG to 1 if
it is not already defined, so that the debugging
facilities are compiled.
Directive: %no-parser
Do not include any C code in the parser file; generate
Directive: %no-lines
Directive: %output="filename"
* the #line directive in Parse.yp for %lexer is prone to be buggy (worst practice)
* Dynamic conflict resolution: test YYSetShift
The case where the default is a reduction is not solved. Investigate
* Further study YAML (class preservation)
* Provide debugging support for standalone modules: currently skipped
* Provide an argument named
'end_of_input' => [ token list: '', NUM, etc. ]
to be used in YYParse. When an error is found, if the token is in the set 'end_of_input' the input
will be accepted if in the accepting state
* There is a bug in eyapp with reduce-reduce conflicts?! I am not sure of this
with reduce-reduce conflicts. In at least one case is reducing by the second production rule instead
of the first. It appears in the C++ parser (t/99Cplusplusgrammar.t). Confirm
* There is a bug with &give_rhs_name in Parse::Eyapp::Grammar and lists *, +, etc. see the example
examples/debuggintut/SemanticInfoInTokens.eyp
Not always a syntactic variable is a valid identifier: e.g. PLUS-1
Probably, it is better that the method "tokennames" give names to the syntactic variables
* str method to produce GraphViz, latex and/or other formats
* Treereg EXPORT and EXPORT_OK
* Language generation example examples/generator
-- Return a Test::LectroTest::Generator?
* Check that all the production names (given with %name directives) are different. May be not
* Combine Flex and eyapp
* Grammar reuse, grammars as subroutines/methods: a grammar subroutine for variable B
is implemented via
B -> FIRST(B) { CALL PARSER FOR B}
The sublanguage is a set of tokens from the superlanguage perspective. The token attribute is the value
returned by the parser. The lexer must be changed to finish when a FOLLOW(B) is found. Line number
and input require special consideration
* Introduce the directive %input FILE. We need to control line numbers and file names for this.
May be we need a full preprocessor for this
* When using option %delegateaction we must check that the client must have
methods with names the names of the production rules (see example in the Simon's book
Advanced programming for this). In fact is a case of "abstract method"
* There is a generic 'Error' routine: provide it
* Not sure: eyappoption -t will generate a 'tag' file
* target 'version' in Makefile.PL: make a vim script
* PUBLIC_PLACE implies to copy to a remote machine (nereida)
* Implement the directive
%package Math::Calc
(like the option -m of eyapp)
* common examples generated through the text template. Idem MANIFEST
* Lexical analyzer seems to be in a mess: line count seem full of bugs
* delete_method isn't documented
* Document generate, new, etc. in treeregexp
* Add to Parse::Eyapp::Base function
MakeHierarchy('BinaryOp(PLUS, TIMES, MINUS, DIV) Operation(BinaryOp)')
that will build a inheritance hierarchy
* Unify: to support trees and not only dags a symbol table will be helpful
* Parse::Eyapp::Node->new: Probably s.t. like
TUTU(TITI({ $[0]->{chu} = 4 }) {$_[0]->{k} = "hello"})
will be better to initialize the attributes
* Document Parse::Eyapp::Base
- Distribution
* Attributed Grammars
* ABSTRACT
* Output files .pm generated by the tests. Don't
* numchildren method
* token definitions %token NUM = qr{\d+}
It can be used later in the lexer as
return ('NUM', $1) if m{($NUM)};
That is the $NUM variable will be declared. Scope?
***** Reorganize the translation of Treeregexp: the code in the "and" and =>
clauses will be a method of the classes of the nodes matching the treeregexp
- $treereg->classes computes the classes. doing nothing yet
- @{$methods{$_}} contain the classes
- Methods Parse::Eyapp::Treeregexp::REGEXP_INNER::classes included
- Factorize
* bug in notation x.y and $x
* Modifiy Parse::Eyapp::Driver::_Parse
If the state is a conflict state execute the conflict handlers for that state:
We need then a hash whose keys will be the conflictive states:
{
state => {
handlername => {
code => conflict handler sub,
actions => { tokens => [a,b,c], shift 2, reduce 4}
}
...
}
...
}
* No syntactic auxiliary variable will be introduced in the %PREC positions.
* The arguments for the 'code' handler will be the parser itself and the conflict name.
DB<3> x $self->{CONFLICTHANDLERS}
0 HASH(0x100b57f30)
'ToN' => HASH(0x100b57eb0)
'.EMPTY' => ARRAY(0x100b43c90)
empty array
'explorerline' => 12
'line' => 5
'production' => HASH(0x100b57de0)
'-6' => ARRAY(0x100b57d80)
0 1
'-7' => ARRAY(0x100b57db0)
0 1
'states' => ARRAY(0x100b57e80)
0 HASH(0x100b57e50)
8 => ARRAY(0x100b57e20)
0 ','
'total' => 2
'totalviewpoint' => 1
* Modify the LR parser algorithm:
my($self)=shift;
my($rules,$states,$error) = @$self{ 'RULES', 'STATES', 'ERROR' };
my($errstatus,$nberror,$token,$value,$stack,$check,$dotpos) = ...
$$errstatus=0; $$nberror=0; ($$token,$$value)=(undef,undef);
@$stack=( [ 0, undef, ] ); $$check='';
while(1) {
my($actions,$act,$stateno);
$stateno=$$stack[-1][0];
$self->executeconflicthandlers($stateno) if ($self->is_conflictive($stateno)); #!!!!
$actions=$$states[$stateno];
....
* Option -v in eyapp will produce a navigable grammar.output.html version of the output
file
* EBNF ors must be implemented
* Artículos:
* OOPLAS
* reescribir con nueva sintáxis TOPLAS y el comsys
* Introducir modificaciones de los referees en los articulos de revista: implementacion, pruebas
* conseguir articulo de Ganapanathi "semantic predicates in"
* estudiar e introducir comparaciones
* insistir en que esta implementado y "verificado": describir implementación
* Referees PLDI
-- too vague description
-- motivation for the YYExpected: explain the problem
-- Syntactic analysis that depends on the semantic phase: explain better
-- how the PPCR is better than backtracking
-- more mathematics, proofs,
-- One pass: is not clear to the referee
-- Doubts about the implementation: have been implemented paragraph, see example files in
such, and such addresses
-- Comparisons with other parsers: Regexp::Grammars, Marpa: execution times, etc.
Complexity.
-- typos, wording
-- two different articles
* optimize generated lexer for contextual tokens (Expected must be called only once)
* debugging: break points, call actual debugger inside semantic actions, etc.
* Trying to supress the generation of conflict rules when there is no conflict (during nested
auxiliary compilation). Count the number of appearances of %PREC conflict . If it is greater
than two there is a true conflict. Otherwise the conflict symbols can be ignored.
There is however a problem. The translation used by Parse.yp performs some premature
semantic actions that insert the conflict symbols as soon as a production is defined.
To solve it we must delay the translation until the whole grammar has been seen.
* Write code to control the right use of %conflict and %PREC
- x $self->{GRAMMAR}{CONFLICTHANDLERS}
0 HASH(0x1009898a0)
'decORexp' => HASH(0x100c5c1e0)
'explorerline' => 14
'line' => 14
'production' => HASH(0x100c50130)
'-12' => ARRAY(0x100c572e0)
0 1
'-5' => ARRAY(0x100c50230)
0 1
'states' => ARRAY(0x100c4bb50)
0 HASH(0x100c5bba0)
28 => ARRAY(0x100c3c5b0)
0 '\')\''
'total' => 2
'totalviewpoint' => 1
DB<8> x $$self{CONFLICTS}{FORCED}
0 HASH(0x100c38800)
'DETAIL' => HASH(0x100c5b630)
28 => HASH(0x100c58030)
'LIST' => ARRAY(0x100c581c0)
0 ARRAY(0x100c581e0)
0 '\')\''
1 '-12'
'TOTAL' => ARRAY(0x100c57d10)
0 empty slot
1 1
'TOTAL' => ARRAY(0x100c57da0)
0 0
1 1
- %PREC in no conflict production or in wrong position
- called for reduce-reduce but it is not a reduce-reduce, etc.
- Every %conflict has to have at least two associated %PRECs
* quiza entre la información guardada en CONFLICTHANDLERS deberían ir las etiquetas simbólicas.
Esto simplificaría YYSetReduce
* Otra API de mas alto nivel sería el caso regexp
%conflict conflictname /regexp/ ruleName
esta se ejecuta en el punto del conflicto
* clean tests deleted. Do it gain?
* examples in MANIFEST: obsolete
* Use %{ ... %} as syntactic sugar for syntactic code actions
* Generation of Several Parsers
- The lexer must be the one of the main parser
* The accept state must change
A$ -> A
A$ -> A.b where b is in the follow of A, means acceptation
Where are the follows?. They have been computed somewhere
* %PREC conflict does not check that a conflict handler has been previously declared
* GLR in eyapp
* eyapp and PPCR in the Wikipedia
* Update documentation
* add eyapp -Y option: equivalent to "%tree bypass"
* Lalr
# substitute End Of Input by DEFAULT for each state
# Will it give a false acceptance?
# I think I shall use an option in eyapp?
# make a test
* Change the API for YYSetReduce and YYSetshift to include the name of the conflict?
* Bug/limitation: When using -m and -o the output file still uses the original grammar name.
It seems more convenient to rename it to the output file suffixed with .output
* Lexical Analyzer generation:
* incremental lexer also influences the way the directive
%lexer
must behave. Because a lexer written by hand can be incremental
* support streams: see lexer in examples/Eyapp/Lhs1.yp
- Parse::Eyapp::Driver::input called as a setter receives an open file handler :
prompt!
- if the input string is finished it reads from the file handler (controled by $/)
otherwise returns the current string
* error line numbers-> source code
* lexer must support a sub name
%lexer mylexer
* Document:
=head3 $parser->input YYInput
=head3 $parser->YYInputFile
=head3 $parser->YYPrompt
=head3 $parser->main
=head3 $parser->Run
* %tree:
- info clause (autobuild ::info for TERMINAL)
- naming scheme clause
* Tests
- tests all examples/*/*yp for syntax checking
- test for yyfilename
- Add tests to check .output tables: actions and gotos against yacc.
I am not confident that some modifications I introduced in Grammar.pm were error free
- Automatic naming scheme: tests!
- Some rules can be optimized away. Then the name dissapears!!
- tests with syntax errors
- Debug mode and standalone works?
* Debugging: set breakpoints in the code from yydebug=> 1
* What if the input comes from a stream? then $parser->input is undef
Shall I support interactive streams?
- YYLookaheads assumes that input comes from $parser->input
- The default lexer assumes that input comes from $parser->input
Change it to use an input stack
* PPCR:
- The whole implementation approach seems wrong.
Make actions an array. Associate the handler with the state.
That allows a syntax like:
%conflict handler {... }
%shift-reduce prod1 [token1, token2, ...] prod2 [tokens] handler
%reduce-reduce prod3 [token1, token2, ...] prod4 [tokens] otherhandler
....
- There is a bug: I have to check that reductions in PPCR only occur inside the conflictive states
- Warning Grammar.pm
sub Parse::Eyapp::Grammar::conflictHandlers rewrite! or use Data::Dumper forever?
* %package directive: equivalent to eyapp -m option
* -%modulino directive
%modulino perlinterpreter prompt
Equivalent to eyapp -b 'perlinterpreter' and __PACKAGE__->main(prompt)
%modulino '/usr/local/lib/perl/5.10.1/bin/perl5.10.1' 'Try input "type e = (x);": '
avoid repetitive __PACKAGE->main when using options -C and -b
* Article: rewrite, both research and divulgation
* Documentation
- Simpler intro
- Separate parsing from tree transformation
* EBNF (a|b|c)*
- eyapp option -TREE: same as %tree
currently only give_rhs is used, but it has to be parametrized
* - eyapp option namingscheme
- eyapp option -start (defines the start symbol)
Eyapp -> Lalr -> Grammar -> Options
`----> Parse
- intermediate actions in give_rhs_name? check, test
* Error message when a conflict handler isn't used is inaccurate:
~/LEyapp/examples/debuggingtut$ eyapp dynamicgrammar.eyp
Useless non-terminals:
rangeORenum, declared line 10
Useless rules:
rangeORenum ->
1 shift/reduce conflict and 0 reduce/reduce conflicts
* Line numbers seem to be in a buggy state.
There is also a "bug" with tokenline and the static attribute approach: tokenline
attribute is defined to 1 while the static/attribute approach searches while "undef"
* Provide a module providing regexp for most common non destructive lexing: NUMs, IDs, etc
use Parse::Eyapp::LexSupport qw{INT PERLWS ID};
Created by "new" so that $self is in scope
$INT = qr{\G(\d+)}gc; $WS = qr{{\G(\s*) (?{ $self->tokenline($1 =~ tr{\n}{}) }) }gxc; etc.
tokens can be "defined" with:
%token DOTDOT /(\.\.)/
%token INT /(\d+)/
The parser will define $DOTDOT = qr{/\G(\.\.)/;
and then:
PERLWS(); # No need to update the line number
m{\Gtype\b}gic and return ('TYPE', 'TYPE'); # old way
m{$ID}gc and return ('ID', $1);
INT() and return ('NUM', $1);
DOTDOT() and return ('DOTDOT', $DOTDOT);
m{\G(.)}gc and return ($1, $1);
* Parse::Eyapp::Driver::main
- option to send output to a file
* bison directives:
Directive: %debug
In the parser file, define the macro YYDEBUG to 1 if
it is not already defined, so that the debugging
facilities are compiled.
Directive: %no-parser
Do not include any C code in the parser file; generate
Directive: %no-lines
Directive: %output="filename"
* the #line directive in Parse.yp for %lexer is prone to be buggy (worst practice)
* Dynamic conflict resolution: test YYSetShift
The case where the default is a reduction is not solved. Investigate
* Further study YAML (class preservation)
* Provide debugging support for standalone modules: currently skipped
* Provide an argument named
'end_of_input' => [ token list: '', NUM, etc. ]
to be used in YYParse. When an error is found, if the token is in the set 'end_of_input' the input
will be accepted if in the accepting state
* There is a bug in eyapp with reduce-reduce conflicts?! I am not sure of this
with reduce-reduce conflicts. In at least one case is reducing by the second production rule instead
of the first. It appears in the C++ parser (t/99Cplusplusgrammar.t). Confirm
* There is a bug with &give_rhs_name in Parse::Eyapp::Grammar and lists *, +, etc. see the example
examples/debuggintut/SemanticInfoInTokens.eyp
Not always a syntactic variable is a valid identifier: e.g. PLUS-1
Probably, it is better that the method "tokennames" give names to the syntactic variables
* str method to produce GraphViz, latex and/or other formats
* Treereg EXPORT and EXPORT_OK
* Language generation example examples/generator
-- Return a Test::LectroTest::Generator?
* Check that all the production names (given with %name directives) are different. May be not
* Combine Flex and eyapp
* Grammar reuse, grammars as subroutines/methods: a grammar subroutine for variable B
is implemented via
B -> FIRST(B) { CALL PARSER FOR B}
The sublanguage is a set of tokens from the superlanguage perspective. The token attribute is the value
returned by the parser. The lexer must be changed to finish when a FOLLOW(B) is found. Line number
and input require special consideration
* Introduce the directive %input FILE. We need to control line numbers and file names for this.
May be we need a full preprocessor for this
* When using option %delegateaction we must check that the client must have
methods with names the names of the production rules (see example in the Simon's book
Advanced programming for this). In fact is a case of "abstract method"
* There is a generic 'Error' routine: provide it
* Not sure: eyappoption -t will generate a 'tag' file
* target 'version' in Makefile.PL: make a vim script
* PUBLIC_PLACE implies to copy to a remote machine (nereida)
* Implement the directive
%package Math::Calc
(like the option -m of eyapp)
* common examples generated through the text template. Idem MANIFEST
* Lexical analyzer seems to be in a mess: line count seem full of bugs
* delete_method isn't documented
* Document generate, new, etc. in treeregexp
* Add to Parse::Eyapp::Base function
MakeHierarchy('BinaryOp(PLUS, TIMES, MINUS, DIV) Operation(BinaryOp)')
that will build a inheritance hierarchy
* Unify: to support trees and not only dags a symbol table will be helpful
* Parse::Eyapp::Node->new: Probably s.t. like
TUTU(TITI({ $[0]->{chu} = 4 }) {$_[0]->{k} = "hello"})
will be better to initialize the attributes
* Document Parse::Eyapp::Base
- Distribution
* Attributed Grammars
* ABSTRACT
* Output files .pm generated by the tests. Don't
* numchildren method
* token definitions %token NUM = qr{\d+}
It can be used later in the lexer as
return ('NUM', $1) if m{($NUM)};
That is the $NUM variable will be declared. Scope?
***** Reorganize the translation of Treeregexp: the code in the "and" and =>
clauses will be a method of the classes of the nodes matching the treeregexp
- $treereg->classes computes the classes. doing nothing yet
- @{$methods{$_}} contain the classes
- Methods Parse::Eyapp::Treeregexp::REGEXP_INNER::classes included
- Factorize
* star treeregexp will be supressed *
* Introduce DECLARATION* in treeregexes
* "previous" clause for YATW objects when using "bud". The YATW will
be executed before visiting the children. An example of the advantage
of doing this is to use "bud" for scope analysis
* %default action must have an scope
* Factorize YYBuildAST and YYActionforT_TX1X2 etc.
There is a problem with TERMINAL::save_attributes ?
bypass, plus and star lists, etc. There is no coherent interface
at the moment
* version numbers
* Warning diagnostic (Yure): ID and 'ID' is prone to error
Non defined variables are considered terminals: see Identifyattr2.yp. Put info in Output
* Better diagnostics: see -v
* Redesign eyapp program structure for documentation (pods): head body legs tail
(or __END__ ?)
* .output in HTML (?)
* Some data structures inside the Parser object are only required
if the %tree or %metatree directives are active. Otimize them away
when they are not used (for example syntax tokens, bypass, etc.).
* Change names of _PLUSLIST, etc. Make a directive to name them
%name * => _STAR + => ... , etc.
It may influence optimizations like the PARENTHESIS one?
* bypass method for YATW and Node objects
* Fix the yyprefix consistency problem between new and YYParse (args)
* What is YYFilename in Driver.pm
* Reference manual. Document functions and methods
-- Sort by alphabetical order
-- User methods and subroutines. Internal documentation
- Treereg Tree grammars and tree walkers
* Uniformize
- Lower case in new_grammar, uppercase in Treeregexp new, etc.
- Also look at scripts
* Private methods
* Indentation and line numbers.
- Translation Schemes
* Scheme examples: leftrec, etc.
* Tests
- In TS delete the code refs using treeregexp
* Documentation
* Executable transscheme
* Delete args in YATW objects it seems they are no use
* TERMINAL nodes inherit from TS and Node. What if I have two parsers
in the same application?
* Put option %yaml file when using a translation scheme to save the AST
* TS: Don't check syntax during the construction phase?
* Check syntax option in eyapp
* Keep Line Numbers in the tree (for tokens)
* Factorize lists code
* Admitir directivas-especificaciones de si la lista debe ser left-recursive o right-recursive
* Esquema de nombres de las reglas temporales:
Reservar la numeración de las reglas de la gramática original
y usar posiciones adicionales para las reglas temporales.
* Hacer una directiva para la generacion de reglas de error en listas
* Supress when production code use Data::Dumper;
* %packagename %outputfile %numbers, etc.
- Optimizations:
* Driver in C
* BuildAST in C
************ PROJECTS *********************
* Human friendly warnings and errors, conflicts, debugger with breaks, etc.