A Lua parser generator that makes it possible to describe grammars in a PEG syntax. The tool will parse a given input using a provided grammar and if the matching is successful produce an AST as an output with the captured values using Lpeg. If the matching fails, labelled errors can be used in the grammar to indicate failure position, and recovery grammars are generated to continue parsing the input using LpegLabel. The tool can also automatically generate error labels and recovery grammars for LL(1) grammars.
parser-gen is a GSoC 2017 project, and was completed with the help of my mentor @sqmedeiros from LabLua. A blog documenting the progress of the project can be found here.
lua >= 5.1
lpeglabel >= 1.2.0
This function generates a PEG parser from the grammar description.
local pg = require "parser-gen"
grammar = pg.compile(input,definitions [, errorgen, noast])Arguments:
input - A string containing a PEG grammar description. For complete PEG syntax see the grammar section of this document.
definitions - table of custom functions and definitions used inside the grammar, for example {equals=equals}, where equals is a function.
errorgen - EXPERIMENTAL optional boolean parameter(default:false), when enabled generates error labels automatically. Works well only on LL(1) grammars. Custom error labels have precedence over automatically generated ones.
noast - optional boolean parameter(default:false), when enabled does not generate an AST for the parse.
Output:
grammar - a compiled grammar on success, throws error on failure.
If custom error labels are used, the function setlabels allows setting their description (and custom recovery pattern):
pg.setlabels(t)Example table of a simple error and one with a custom recovery expression:
-- grammar rule: " ifexp <- 'if' exp 'then'^missingThen stmt 'end'^missingEnd "
local t = {
missingEnd = "Missing 'end' in if expression",
missingThen = {"Missing 'then' in if expression", " (!stmt .)* "} -- a custom recovery pattern
}
pg.setlabels(t)If the recovery pattern is not set, then the one specified by the rule SYNC will be used. It is by default set to:
SKIP <- %s / %nl -- a space ' ' or newline '\n' character
SYNC <- .? (!SKIP .)*Learn more about special rules in the grammar section.
This operation attempts to match a grammar to the given input.
result, errors = pg.parse(input, grammar [, errorfunction])Arguments:
input - an input string that the tool will attempt to parse.
grammar - a compiled grammar.
errorfunction - an optional function that will be called if an error is encountered, with the arguments desc for the error description set using setlabels(); location indicators line and col; the remaining string before failure sfail and a custom recovery expression trec if available.
Example:
local errs = 0
local function printerror(desc,line,col,sfail,trec)
errs = errs+1
print("Error #"..errs..": "..desc.." before '"..sfail.."' on line "..line.."(col "..col..")")
end
result, errors = pg.parse(input,grammar,printerror)Output:
If the parse is succesful, the function returns an abstract syntax tree containing the captures result and a table of any encountered errors. If the parse was unsuccessful, result is going to be nil.
Also, if the noast option is enabled when compiling the grammar, the function will then produce the longest match length or any custom captures used.
Calculates line and column information regarding position i of the subject (exported from the relabel module).
line, col = pg.calcline(subject, position)Arguments:
subject - subject string
position - position inside the string, for example, the one given by automatic AST generation.
When AST generation is enabled, this function will enable the "node" mode, where only rules tagged with a node prefix will generate AST entries. Must be used before compiling the grammar.
pg.usenodes(value)Arguments:
value - a boolean value that enables or disables this function
The grammar used for this tool is described using a PEG-like syntax, that is identical to the one provided by the re module, with an extension of labelled failures provided by relabel module (except numbered labels). That is, all grammars that work with relabel should work with parser-gen as long as numbered error labels are not used, as they are not supported by parser-gen.
Since a parser generated with parser-gen automatically consumes space characters, builds ASTs and generates errors, additional extensions have been added based on the ANTLR syntax.
The syntax of parser-gen grammars is somewhat similar to regex syntax. The next table summarizes the tools syntax. A p represents an arbitrary pattern; num represents a number ([0-9]+); name represents an identifier ([a-zA-Z][a-zA-Z0-9_]*).defs is the definitions table provided when compiling the grammar. Note that error names must be set using setlabels() before compiling the grammar. Constructions are listed in order of decreasing precedence.
| Syntax | Description |
( p ) | grouping |
'string' | literal string |
"string" | literal string |
[class] | character class |
. | any character |
%name |
pattern defs[name] or a pre-defined pattern |
name | non terminal |
<name> | non terminal |
%{name} | error label |
{} | position capture |
{ p } | simple capture |
{: p :} | anonymous group capture |
{:name: p :} | named group capture |
{~ p ~} | substitution capture |
{| p |} | table capture |
=name | back reference |
p ? | optional match |
p * | zero or more repetitions |
p + | one or more repetitions |
p^num | exactly n repetitions |
p^+num |
at least n repetitions |
p^-num |
at most n repetitions |
p^name | match p or throw error label name. |
p -> 'string' | string capture |
p -> "string" | string capture |
p -> num | numbered capture |
p -> name | function/query/string capture
equivalent to p / defs[name] |
p => name | match-time capture
equivalent to lpeg.Cmt(p, defs[name]) |
& p | and predicate |
! p | not predicate |
p1 p2 | concatenation |
p1 //{name [, name, ...]} p2 | specifies recovery pattern p2 for p1 when one of the labels is thrown |
p1 / p2 | ordered choice |
(name <- p)+ | grammar |
The grammar below is used to match balanced parenthesis
balanced <- "(" ([^()] / balanced)* ")" For more examples check out the re page, see the Tiny parser below or the Lua parser writen with this tool.
Error labels are provided by the relabel function %{errorname} (errorname must follow [A-Za-z][A-Za-z0-9_]* format). Usually we use error labels in a syntax like 'a' ('b' / %{errB}) 'c', which throws an error label if 'b' is not matched. This syntax is quite complicated so an additional syntax is allowed 'a' 'b'^errB 'c', which allows cleaner description of grammars. Note: all errors must be defined in a table using parser-gen.setlabels() before compiling and parsing the grammar.
Non-terminals with names in all capital letters, i.e. [A-Z]+, are considered tokens and are treated as a single object in parsing. That is, the whole string matched by a token is captured in a single AST entry and space characters are not consumed. Consider two examples:
-- a token non-terminal
grammar = pg.compile [[
WORD <- [A-Z]+
]]
res, _ = pg.parse("AA A", grammar) -- outputs {rule="WORD", "AA"}-- a non-token non-terminal
grammar = pg.compile [[
word <- [A-Z]+
]]
res, _ = pg.parse("AA A", grammar) -- outputs {rule="word", "A", "A", "A"}If a token definition is followed by a fragment keyword, then the parser does not build an AST entry for that token. Essentially, these rules are used to simplify grammars without building unnecessarily complicated ASTS. Example of fragment usage:
grammar = pg.compile [[
WORD <- LETTER+
fragment LETTER <- [A-Z]
]]
res, _ = pg.parse("AA A", grammar) -- outputs {rule="WORD", "AA"}Without using fragment:
grammar = pg.compile [[
WORD <- LETTER+
LETTER <- [A-Z]
]]
res, _ = pg.parse("AA A", grammar) -- outputs {rule="WORD", {rule="LETTER", "A"}, {rule="LETTER", "A"}}
When node mode is enabled using pg.usenodes(true) only rules prefixed with a node keyword will generate AST entries:
grammar = pg.compile [[
node WORD <- LETTER+
LETTER <- [A-Z]
]]
res, _ = pg.parse("AA A", grammar) -- outputs {rule="WORD", "AA"}There are two special rules used by the grammar:
The SKIP rule identifies which characters to skip in a grammar. For example, most programming languages do not take into acount any space or newline characters. By default, SKIP is set to:
SKIP <- %s / %nlThis rule can be extended to contain semicolons ';', comments, or any other patterns that the parser can safely ignore.
Character skipping can be disabled by using:
SKIP <- ''This rule specifies the general recovery expression both for custom errors and automatically generated ones. By default:
SYNC <- .? (!SKIP .)*The default SYNC rule consumes any characters until the next character matched by SKIP, usually a space or a newline. That means, if some statement in a program is invalid, the parser will continue parsing after a space or a newline character.
For some programming languages it might be useful to skip to a semicolon or a keyword, since they usually indicate the end of a statement, so SYNC could be something like:
HELPER <- ';' / 'end' / SKIP -- etc
SYNC <- (!HELPER .)* SKIP* -- we can consume the spaces after syncing with them as wellRecovery grammars can be disabled by using:
SYNC <- ''Below is the full code from parsers/tiny-parser.lua:
local pg = require "parser-gen"
local peg = require "peg-parser"
local errs = {errMissingThen = "Missing Then"} -- one custom error
pg.setlabels(errs)
--warning: experimental error generation function is enabled. If the grammar isn't LL(1), set errorgen to false
local errorgen = true
local grammar = pg.compile([[
program <- stmtsequence !.
stmtsequence <- statement (';' statement)*
statement <- ifstmt / repeatstmt / assignstmt / readstmt / writestmt
ifstmt <- 'if' exp 'then'^errMissingThen stmtsequence elsestmt? 'end'
elsestmt <- ('else' stmtsequence)
repeatstmt <- 'repeat' stmtsequence 'until' exp
assignstmt <- IDENTIFIER ':=' exp
readstmt <- 'read' IDENTIFIER
writestmt <- 'write' exp
exp <- simpleexp (COMPARISONOP simpleexp)*
COMPARISONOP <- '<' / '='
simpleexp <- term (ADDOP term)*
ADDOP <- [+-]
term <- factor (MULOP factor)*
MULOP <- [*/]
factor <- '(' exp ')' / NUMBER / IDENTIFIER
NUMBER <- '-'? [0-9]+
KEYWORDS <- 'if' / 'repeat' / 'read' / 'write' / 'then' / 'else' / 'end' / 'until'
RESERVED <- KEYWORDS ![a-zA-Z]
IDENTIFIER <- !RESERVED [a-zA-Z]+
HELPER <- ';' / %nl / %s / KEYWORDS / !.
SYNC <- (!HELPER .)*
]], _, errorgen)
local errors = 0
local function printerror(desc,line,col,sfail,trec)
errors = errors+1
print("Error #"..errors..": "..desc.." on line "..line.."(col "..col..")")
end
local function parse(input)
errors = 0
result, errors = pg.parse(input,grammar,printerror)
return result, errors
end
if arg[1] then
-- argument must be in quotes if it contains spaces
res, errs = parse(arg[1])
peg.print_t(res)
peg.print_r(errs)
end
local ret = {parse=parse}
return retFor input: lua tiny-parser-nocap.lua "if a b:=1" we get:
Error #1: Missing Then on line 1(col 6)
Error #2: Expected stmtsequence on line 1(col 9)
Error #3: Expected 'end' on line 1(col 9)
-- ast:
rule='program',
pos=1,
{
rule='stmtsequence',
pos=1,
{
rule='statement',
pos=1,
{
rule='ifstmt',
pos=1,
'if',
{
rule='exp',
pos=4,
{
rule='simpleexp',
pos=4,
{
rule='term',
pos=4,
{
rule='factor',
pos=4,
{
rule='IDENTIFIER',
pos=4,
'a',
},
},
},
},
},
},
},
},
-- error table:
[1] => {
[msg] => 'Missing Then' -- custom error is used over the automatically generated one
[line] => '1'
[col] => '6'
[label] => 'errMissingThen'
}
[2] => {
[msg] => 'Expected stmtsequence' -- automatically generated errors
[line] => '1'
[col] => '9'
[label] => 'errorgen6'
}
[3] => {
[msg] => 'Expected 'end''
[line] => '1'
[col] => '9'
[label] => 'errorgen4'
}
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent
