I'm writing a lexer and parser in ocamllex and ocamlyacc as follows. function_name and table_name are same regular expression, i.e., a string containing only english alphabets. The only way to determine if a string is function_name or table_name is to check its surroundings. For example, if such a string is surrounded by [ and ], then we know that it is a table_name. Here is the current code:
In lexer.mll,
... ...
let function_name = ['a'-'z' 'A'-'Z']+
let table_name = ['a'-'z' 'A'-'Z']+
rule token = parse
| function_name as s { FUNCTIONNAME s }
| table_name as s { TABLENAME s }
... ...
In parser.mly:
... ...
main:
| LBRACKET TABLENAME RBRACKET { Table $2 }
... ...
As I wrote | function_name as s { FUNCTIONNAME s } before | table_name as s { TABLENAME s }, the above code failed to parse [haha]; it firstly considered haha as a function_name in the lexer, then it could not find any corresponding rule for it in the parser. If it could consider haha as a table_name in the lexer, it would match [haha] as a table in the parser.
One workaround for this is to be more precise in the lexer. For example, we define let table_name_with_brackets = '[' ['a'-'z' 'A'-'Z']+ ']' and | table_name_with_brackets as s { TABLENAMEWITHBRACKETS s } in the lexer. But, I would like to know if there is any other options. Is it not possible to make lexer and parser work together to determine the tokens and the reduction?
You should avoid trying to get the lexer to do the parser's work. The lexer should just identify lexemes; it should not try to figured out where a lexeme fits into the syntax. So in your (simplified) example, there should be only one lexical type, name. The parser will figure it out from there.
But it seems, from the comments, that in the unsimplified original, the two patterns are overlapping rather than identical. That's more annoying, although it's only slightly more complicated. Basically, you need to separate out the common pattern as one lexical type, and then add the additional matches as one or two other lexical types (depending on whether or not one pattern is a strict superset of the other).
That might not be too difficult, depending on the precise relationship between the two patterns. You might be able to find a very simple solution by writing the patterns in the correct order, for example, because of the longest match rule:
If several regular expressions match a prefix of the input, the “longest match” rule applies: the regular expression that matches the longest prefix of the input is selected. In case of tie, the regular expression that occurs earlier in the rule is selected.
Most of the time, that's all it takes: first define the intersection of the two patterns as a based lexeme, and then add the full lexical patterns of each contextual type to provide additional matches. Your parser will then have to match name | function_name in one context and name | table_name in the other context. But that's not too bad.
Where it will fail is when an input stream cannot be unambiguously divided in lexemes. For example, suppose that in a function context, a name could include a ? character, but in a table context the ? is a valid postscript operator. In that case, you have to actively prevent foo? from being analysed as a single token in the table context, which means that the lexer does have to be aware of parser context.
Related
Sometimes I get a bit confused between a lexing rule vs. a parsing rule, and there's been a nice thread on it here. For example in the following:
value
: string CAST_OPERATOR type
;
string
: S_QUOTE STRING_VALUE S_QUOTE
;
# <-- what is this?
type
: 'date' | 'string'
;
STRING_VALUE
: [a-zA-Z0-9-]+
;
CAST_OPERATOR
: '::'
;
For the type -- this is either the string (or character stream) date or string. Should that be a lexing rule or a parsing rule? I suppose I could break it down even more into:
type
: DATE_TYPE | STRING_TYPE
;
DATE_TYPE
: 'date'
;
STRING_TYPE
: 'string'
;
But still I'm not quite sure which of the above is preferable, and why it would be so. The first two rules -- value and string seem clear to me to be parsing rules -- and the last two rules -- STRING_VALUE and CAST_OPERATOR seem clear to me to be lexing rules (only by intuition though, I could not give a proper explanation). So why would the type be one way or the other?
Literally the only practical difference I've found is that a lexing rule can include a character class and a parsing rule cannot.
Update: I suppose another thing is a lexing rule is terminal, it won't provide any subdivision of parts. For example in the following we can break down $55 into $ and 55:
But if we set the cost as a lexing rule, it will not break it down any further:
So basically a lexing rule is atomic and terminal, whereas a parsing rule is more like a molecule that consists of various parts (atoms) that can be seen within it. Is that a good description/understanding of it?
Your "Update" is on the right track. That's a definite distinction.
You also need to understand the ANTLR pipeline. I.e. that the stream of characters is processed by the Lexer rules to produce a stream of tokens (atoms, in you analogy). It does not do that with recursive descent rule matching, but rather attempts to match you input against all of the Lexer rules. Where:
The rule that matches the longest sequence of input characters will "win"
In the event that multiple Lexer rules match the same length character sequence, then the rule that occurs first will "win"
Once you've got you stream of "atoms" (aka Tokens), then ANTLR uses the parser rules (recursively from the start rule) to try to match sequences of tokens.
I am writing a lexer/parser for a language that allows abbreviations (and globs) for its keywords. And, I am trying to determine the best way to do it.
And one thought that occurs to me, is to insert a phase between the lexer and the parser, where the lexer recognizes the general class, e.g. is this a "command name" or is it an "option" and then passes those general tokens to a second phase which does further analysis and recognizes which command name it is and passes that on as the token type to the parser.
It will make the parser simple. I will only have to deal with well formed command names. Every token will be clear what it means.
It will keep the lexer simple. It will only have to divide things into classes. This is a simple name. This is a glob. This is an option name (starts with a dash).
The phase is the middle will also be relatively simple. The simple name (and option forms) will only have to deal with strings. The glob form can use standard glob techniques to match the glob against the legal candidates, which are in the tables for the simple names and options.
The question is how to insert that phase into ANTLR, so that I call the lexer and it creates tokens and the intermediate phase massages them and then the parser gets the tokens the intermediate phase has categorized.
Is there a known solution for this?
Something like:
lexer grammar simple
letter: [A-Z][a-z];
digit: [0-9];
glob-char: [*?];
name: letter (letter | digit)*;
option: '-'name;
glob: (glob-char|letter)(glob-char|letter|digit)*;
glob-option: '-'glob;
filter grammar name;
end: 'e' | 'end';
generate: 'ge' | 'generate';
goto: 'go' | 'goto';
help: 'h' | 'help';
if: 'i' | 'if';
then: 't' | 'then';
parser grammar simple;
The user (programmer writing the language I am parsing) need to be to write
g*te and have if match generate.
The phase between the lexer and the parser when it sees a glob needs to look at the glob (and the list of keywords) and see if only one of them matches the glob and if so, return that keyword. The stuff I listed in the "filter grammar" is the stuff that builds the list of keywords globs can match. I have found code on the web that matches globs to a list of names. That part isn't hard.
And, I've since found in the ANTLR doc how to run arbitrary code on matching a token and how to change the resulting tokens type. (See my answer.)
It looks like you can use lexerCustomActions to achieve the desired effect. Something like the following.
in your lexer:
GLOB: [-A-Za-z0-9_.]* '*' [-A-Za-z0-9_.*]* { setType(lexGlob(getText())); }
in your Java (or whatever language you are using code):
void int lexGlob(String origText()) {
return xyzzy; // some code that computes the right kind of token type
}
Using ANTLR 3, my lexer has rule
SELECT_ASSIGN:
'SELECT' WS+ IDENTIFIER WS+ 'ASSIGN' WS+ (('TO'|'USING') WS+)?
using this these match correctly
SELECT VAR1 ASSIGN TO
SELECT VAR1 ASSIGN USING
and this also matches
SELECT VAR1 ASSIGN FOO
However this does not match
SELECT VAR1 ASSIGN TWO
Whereas I have marked TO|USING as optional in the rule.
From generated Java code I see...
When lexer notices T of TWO, it goes to match('TO')
but since does not find O after T
then generates failure.... and returns all the way from the rule -- hence not matching it.
How do I get my lexer rule to match, when input has word with chars starting with suffixed optional part of the rule
Basically I want my rule to match this also (beside what it already matches - as lised at the start):
SELECT VAR1 ASSIGN TWO
Kindly suggest how I approach/resolve this situation.
NOTE:
Such rules are recommended in the parser - But I have this in lexer - because I do not want to parse the entire input by the parser, and want to parse only content of interest. So using such rules in lexer, I locate sections which I really want to parse by the parser.
UPDATE 1
I could circumvent this problem by making 2 rules, like so:
SELECT_ASSIGN_USING_TO
: tok='SELECT' WS+ name=IDENTIFIER WS+ 'ASSIGN' WS+ ('USING'|'TO')
SELECT_ASSIGN
: tok='SELECT' WS+ name=IDENTIFIER WS+ 'ASSIGN'
But is it possible to do the desired in one lexer rule?
An approach to get this in one rule, suggested by my senior - use syntactic predicate
SELECT_ASSIGN
: tok='SELECT' WS+ name=IDENTIFIER WS+ 'ASSIGN'
(
(WS+ ('TO'|'USING') WS+)=> (WS+ ('TO'|'USING') WS+)
| (WS+)
)
Tokens match a complete char sequence or none. It cannot match partially and the grammar rule determines which exactly. You cannot expect a rule for TO to match TWO. If you want TWO to match too you have to add it to your lexer rule.
A few notes here:
The solution your "senior" gave you makes no sense at all. A
syntactic predicate is a kinda lookahead to guide the parser in case
of ambiquities. There are no ambiquities involved here.
Writing
the entire SELECT_ASSIGN rule as a lexer rule is very uncommon and
not flexible. A lexer rule should not be used for entire sentences,
but only for a small set of characters to find tokens to assign them
a type (usually elementary structures of a language like string,
number, comment etc.).
ANTLR3 is totally outdated and I wonder why this is still used in your class. ANTLR4 is out since 5 years and should be the choice for any new project.
"end" { return 'END'; }
...
0[xX][0-9a-fA-F]+ { return 'NUMBER'; }
[A-Za-z_$][A-Za-z0-9_$]* { return 'IDENT'; }
...
Call
: IDENT ArgumentList
{{ $$ = ['CallExpr', $1, $2]; }}
| IDENT
{{ $$ = ['CallExprNoArgs', $1]; }}
;
CallArray
: CallElement
{{ $$ = ['CallArray', $1]; }}
;
CallElement
: CallElement "." Call
{{ $$ = ['CallElement', $1, $3]; }}
| Call
;
Hello! So, in my grammar I want "res.end();" to not detect end as a keyword, but as an ident. I've been thinking for a while about this one but couldn't solve it. Does anyone have any ideas? Thank you!
edit: It's a C-like programming language.
There's not quite enough information in the question to justify the assumptions I'm making here, so this answer may be inexact.
Let's suppose we have a somewhat Lua-like language in which a.b is syntactic sugar for a["b"]. Furthermore, since the . must be followed by a lexical identifier -- in other words, it is never followed by a syntactic keyword -- we'd like to inhibit keyword recognition in this context.
That's a pretty simple rule. It's simple enough that the lexer could implement it without any semantic information at all; all that it says is that the token which follows a . must be an identifier. In this context, keywords should be treated as identifiers, and anything else other than an identifier is an error.
We can do this with start conditions. Specifically, we define a start condition which is only used after a . token:
%x selector
%%
/* White space and comment rules need to explicitly include
* the selector condition
*/
<INITIAL,selector>\s+ ;
/* Other rules, including keywords, are unmodified */
"end" return "END";
/* The dot rule triggers a new start condition */
"." this.begin("selector"); return ".";
/* Outside of the start condition, identifiers don't change state. */
[A-Za-z_]\w* yylval = yytext; return "ID";
/* Only identifiers are valid in this start condition, and if found
* the start condition is changed back. Anything else is an error.
*/
<selector>[A-Za-z_]\w* yylval = yytext; this.popState(); return "ID";
<selector>. parse_error("Expecting identifier");
Modify your parser, so it always knows what it is expecting to read next (that will be some set of tokens, you can compute this using the notion of First(x) for x being any nonterminal).
When lexing, have the lexer ask the parser what set of tokens it expects next.
Your keywork reconizer for 'end' asks the parser, and it either ways "expecting 'end'" at which pointer the lexer simply hands on the 'end' lexeme, or it says "expecting ID" at which point it hands the parser an ID with name text "end".
This may or may not be convenient to get your parser to do. But you need something like this.
We use a GLR parser; our parser accepts multiple tokens in the same place. Our solution is to generate both the 'end' keyword and and the identifier with text "end" and shove them both into the GLR parser. It can handle local ambiguity; the multiple parses caused by this proceed until the parser with the wrong assumption encounters a syntax error, and then it just vanishes, by fiat. The last standing parser is the one with the right set of assumptions. This scheme is somewhat like the first one, just that we hand the parser the choices and it decides rather than making the lexer decide.
You might be able to send your parser a "two-interpretation" lexeme, e.g., a keyword-in-context lexeme, which in essence claims it it both a keyword and/or an identifier. With a single token lookahead internally, the parser can likely decide easily and restamp the lexeme. Not as general as the GLR solution, but probably works in a lot of cases.
If someone would clear my mind from the confusion behind look-ahead relation to tokenizing involving greery/non-greedy matching i'd be more than glad. Be ware this is a slightly long post because it's following my thought process behind.
I'm trying to write antlr3 grammar that allows me to match input such as:
"identifierkeyword"
I came up with a grammar like so in Antlr 3.4:
KEYWORD: 'keyword' ;
IDENTIFIER
:
(options {greedy=false;}: (LOWCHAR|HIGHCHAR))+
;
/** lowercase letters */
fragment LOWCHAR
: 'a'..'z';
/** uppercase letters */
fragment HIGHCHAR
: 'A'..'Z';
parse: IDENTIFIER KEYWORD EOF;
however it complains about it can never match IDENTIFIER this way, which i don't really understand. (The following alternatives can never be matched: 1)
Basically I was trying to specify for the lexer that try to match (LOWCHAR|HIGHCHAR) non-greedy way so it stops at KEYWORD lookahead. What i've read so far about ANTLR lexers that there supposed to be some kind of precedence of the lexer rules. If i specify KEYWORD lexer rule first in the lexer grammar, any lexer rules that come after shouldn't be able to match the consumed characters.
After some searching I understand that problem here is that it can't tokenize the input the right way because for example for input: "identifierkeyword" the "identifier" part comes first so it decides to start matching the IDENTIFIER rule when there is no KEYWORD tokens matched yet.
Then I tried to write the same grammar in ANTLR 4, to test if the new run-ahead capabilities can match what i want, it looks like this:
KEYWORD: 'keyword' ;
/** lowercase letters */
fragment LOWCHAR
: 'a'..'z';
/** uppercase letters */
fragment HIGHCHAR
: 'A'..'Z';
IDENTIFIER
:
(LOWCHAR|HIGHCHAR)+?
;
parse: IDENTIFIER KEYWORD EOF;
for the input: "identifierkeyword" it produces this error:
line 1:1 mismatched input 'd' expecting 'keyword'
it matches character 'i' (the very first character) as an IDENTIFIER token, and then the parser expects a KEYWORD token which he doesn't get this way.
Isn't the non-greedy matching for the lexer supposed to match till any other possibility is available in the look ahead? Shouldn't it look ahead for the possibility that an IDENTIFIER can contain a KEYWORD and match it that way?
I'm really confused about this, I have watched the video where Terence Parr introduces the new capabilities of ANTLR4 where he talks about run-ahead threads that watch for all "right" solutions till the end while actually matching a rule. I thought it would work for Lexer rules too, where a possible right solution for tokenizing input "identifierkeyword" is matching IDENTIFIER: "identifier" and matching KEYWORD: "keyword"
I think I have lots of wrongs in my head about non-greedy/greedy matching. Could somebody please explain me how it works?
After all this I've found a similar question here: ANTLR trying to match token within longer token and made a grammar corresponding to that:
parse
:
identifier 'keyword'
;
identifier
:
(HIGHCHAR | LOWCHAR)+
;
/** lowercase letters */
LOWCHAR
: 'a'..'z';
/** uppercase letters */
HIGHCHAR
: 'A'..'Z';
This does what I want now, however I can't see why I can't change the identifier rule to a Lexer rule and LOWCHAR and HIGHCHAR to fragments.
A Lexer doesn't know that letters in "keyword" can be matched as an identifier? or vice versa? Or maybe it is that rules are only defined to have a lookahead inside themselves, not all possible matching syntaxes?
The easiest way to resolve this in both ANTLR 3 and ANTLR 4 is to only allow IDENTIFIER to match a single input character, and then create a parser rule to handle sequences of these characters.
identifier : IDENTIFIER+;
IDENTIFIER : HIGHCHAR | LOWCHAR;
This would cause the lexer to skip the input identifier as 10 separate characters, and then read keyword as a single KEYWORD token.
The behavior you observed in ANTLR 4 using the non-greedy operator +? is similar to this. This operator says "match as few (HIGHCHAR|LOWCHAR) blocks as possible while still creating an IDENTIFIER token". Clearly the fewest number to create the token is one, so this was effectively a highly inefficient way of writing IDENTIFIER to match a single character. The reason the parse rule failed to handle this is it only allows a single IDENTIFIER token to appear before the KEYWORD token. By creating a parser rule identifier like I showed above, the parser would be able to treat sequences of IDENTIFIER tokens (which are each a single character), as a single identifier.
Edit: The reason you get the message "The following alternatives can never be matched..." in ANTLR 3 is the static analysis has determined that the positive closure in the rule IDENTIFIER will never match more than 1 character because the rule will always be successful with exactly 1 character.