Lexer Generator and Parser Generator as a Macro Library in Nim.
With nimly, you can make lexer/parser by writing definition
in formats like lex/yacc.
nimly generates lexer and parser by using macro in compile-time,
so you can use nimly not as external tool of your program but as a library.
niml is a macro to generate a lexer.
macro niml makes a lexer.
Almost all part of constructing a lexer is done in compile-time.
Example is as follows.
## This makes a LexData object named myLexer.
## This lexer returns value with type ``Token`` when a token is found.
niml myLexer[Token]:
r"if":
## this part converted to procbody.
## the arg is (token: LToken).
return TokenIf()
r"else":
return TokenElse()
r"true":
return TokenTrue()
r"false":
return TokenFalse()
## you can use ``..`` instead of ``-`` in ``[]``.
r"[a..zA..Z\-_][a..zA..Z0..9\-_]*":
return TokenIdentifier(token)
## you can define ``setUp`` and ``tearDown`` function.
## ``setUp`` is called from ``open``, ``newWithString`` and
## ``initWithString``.
## ``tearDown`` is called from ``close``.
## an example is ``test/lexer_global_var.nim``.
setUp:
doSomething()
tearDown:
doSomething()Meta charactors are as following:
\: escape character.: match with any charactor[: start of character class|: means or(: start of subpattern): end of subpattern?: 0 or 1 times quantifier*: 0 or more times quantifire+: 1 or more times quantifire{:{n,m}is n or more and m or less times quantifire
In [], meta charactors are as following
\: escape character^: negate character (only in first position)]: end of this class-: specify character range (..can be used instead of this)
Each of followings is recognized as character set.
\d:[0..9]\D:[^0..9]\s:[ \t\n\r\f\v]\S:[^ \t\n\r\f\v]\w:[a..zA..Z0..9_]\w:[^a..zA..Z0..9_]
nimy is a macro to generate a LALR(1) parser.
macro nimy makes a parser.
Almost all part of constructing a parser is done in compile-time.
Example is as follows.
## This makes a LexData object named myParser.
## first cloud is the top-level of the BNF.
## This lexer recieve tokens with type ``Token`` and token must have a value
## ``kind`` with type enum ``[TokenTypeName]Kind``.
## This is naturally satisfied when you use ``patty`` to define the token.
nimy myParser[Token]:
## the starting non-terminal
## the return type of the parser is ``Expr``
top[Expr]:
## a pattern.
expr:
## proc body that is used when parse the pattern with single ``expr``.
## $1 means first position of the pattern (expr)
return $1
## non-terminal named ``expr``
## with returning type ``Expr``
expr[Expr]:
## first pattern of expr.
## ``LPAR`` and ``RPAR`` is TokenKind.
LPAR expr RPAR:
return $2
## second pattern of expr.
## ``PLUS`` is TokenKind.
expr PLUS expr
return $2You can use following EBNF functions:
XXX[]: Option (0 or 1XXX). The type isseq[xxx]wherexxxis type ofXXX.XXX{}: Repeat (0 or moreXXX). The type isseq[xxx]wherexxxis type ofXXX.
Example of these is in next section.
tests/test_readme_example.nim is an easy example.
import unittest
import patty
import strutils
import nimly
## variant is defined in patty
variant MyToken:
PLUS
MULTI
NUM(val: int)
DOT
LPAREN
RPAREN
IGNORE
niml testLex[MyToken]:
r"\(":
return LPAREN()
r"\)":
return RPAREN()
r"\+":
return PLUS()
r"\*":
return MULTI()
r"\d":
return NUM(parseInt(token.token))
r"\.":
return DOT()
r"\s":
return IGNORE()
nimy testPar[MyToken]:
top[string]:
plus:
return $1
plus[string]:
mult PLUS plus:
return $1 & " + " & $3
mult:
return $1
mult[string]:
num MULTI mult:
return "[" & $1 & " * " & $3 & "]"
num:
return $1
num[string]:
LPAREN plus RPAREN:
return "(" & $2 & ")"
## float (integer part is 0-9) or integer
NUM DOT[] NUM{}:
result = ""
# type of `($1).val` is `int`
result &= $(($1).val)
if ($2).len > 0:
result &= "."
# type of `$3` is `seq[MyToken]` and each elements are NUM
for tkn in $3:
# type of `tkn.val` is `int`
result &= $(tkn.val)
test "test Lexer":
var testLexer = testLex.newWithString("1 + 42 * 101010")
testLexer.ignoreIf = proc(r: MyToken): bool = r.kind == MyTokenKind.IGNORE
var
ret: seq[MyTokenKind] = @[]
for token in testLexer.lexIter:
ret.add(token.kind)
check ret == @[MyTokenKind.NUM, MyTokenKind.PLUS, MyTokenKind.NUM,
MyTokenKind.NUM, MyTokenKind.MULTI,
MyTokenKind.NUM, MyTokenKind.NUM, MyTokenKind.NUM,
MyTokenKind.NUM, MyTokenKind.NUM, MyTokenKind.NUM]
test "test Parser 1":
var testLexer = testLex.newWithString("1 + 42 * 101010")
testLexer.ignoreIf = proc(r: MyToken): bool = r.kind == MyTokenKind.IGNORE
var parser = testPar.newParser()
check parser.parse(testLexer) == "1 + [42 * 101010]"
testLexer.initWithString("1 + 42 * 1010")
parser.init()
check parser.parse(testLexer) == "1 + [42 * 1010]"
test "test Parser 2":
var testLexer = testLex.newWithString("1 + 42 * 1.01010")
testLexer.ignoreIf = proc(r: MyToken): bool = r.kind == MyTokenKind.IGNORE
var parser = testPar.newParser()
check parser.parse(testLexer) == "1 + [42 * 1.01010]"
testLexer.initWithString("1. + 4.2 * 101010")
parser.init()
check parser.parse(testLexer) == "1. + [4.2 * 101010]"
test "test Parser 3":
var testLexer = testLex.newWithString("(1 + 42) * 1.01010")
testLexer.ignoreIf = proc(r: MyToken): bool = r.kind == MyTokenKind.IGNORE
var parser = testPar.newParser()
check parser.parse(testLexer) == "[(1 + 42) * 1.01010]"nimble install nimly
Now, you can use nimly with import nimly.
During compiling lexer/parser, you can encounter errors with interpretation requires too many iterations.
You can avoid this error to use the compiler option maxLoopIterationsVM:N
which is available since nim v1.0.6.
See #11 to detail.
- Fork this
- Create new branch
- Commit your change
- Push it to the branch
- Create new pull request
See changelog.rst.
You can use nimldebug and nimydebug as a conditional symbol
to print debug info.
example: nim c -d:nimldebug -d:nimydebug -r tests/test_readme_example.nim