summaryrefslogtreecommitdiff
path: root/vndbapi-msg/src/parser.rs
blob: 135c71d136e18de3974dc00e46a7747a3dbfcb13 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
use serde_json::Value;

use msg::{Filter,Op,Arg,Message};

type Result<T> = ::std::result::Result<T, &'static str>;



/* Parse a JSON value with trailing data. This is a workaround until a proper solution has
 * been implemented: https://github.com/serde-rs/json/issues/183 */
fn parse_json(s: &str) -> Result<(Value, &str)> {
    println!("Parsing JSON: {}", s);
    let mut bytes = 0;
    let val: Value = ::serde_json::Deserializer::from_iter(s.bytes().map(|b| { bytes += 1; Ok(b)} ))
        .into_iter().next()
        .ok_or("Expected JSON value")?
        .map_err(|_| "Invalid JSON value")?;

    // The JSON deserializer consumes one extra byte for numeric types, subtract that.
    match val { Value::Number(_) => bytes -= 1, _ => () };

    Ok((val, &s[bytes..]))
}


fn is_ws(c: char) -> bool { c == ' ' || c == '\t' || c == '\r' || c == '\n' }
fn is_filtername(c: char) -> bool { (c >= 'a' && c <= 'z') || c == '_' }
fn is_barestr(c: char) -> bool { (c >= 'a' && c <= 'z') || c == '_' || c == ',' }
fn trim_ws(s: &str) -> &str { s.trim_left_matches(is_ws) }




#[derive(Debug,PartialEq,Clone,Copy)]
enum Token {
    Open,
    Close,
    And,
    Or,
    Expr,
}

pub struct FilterParser<'a> {
    buf: &'a str,
    hasexpr: bool,
}


impl<'a> FilterParser<'a> {
    // Consume any whitespace
    fn conws(&mut self) {
        self.buf = trim_ws(self.buf);
    }

    // Consume the given number of bytes
    fn con(&mut self, bytes: usize) {
        self.buf = &self.buf[bytes..];
    }

    fn token_expr(&mut self) -> Result<Filter> {
        let name: String = self.buf.chars().take_while(|&c| is_filtername(c)).collect();
        if name.len() == 0 {
            return Err("Invalid token");
        }
        self.con(name.len());
        self.conws();

        let op = parse_op(self.buf).ok_or("Expected comparison operator")?;
        self.con(op.as_str().len());
        self.conws();

        let val = parse_json(self.buf)?;
        self.buf = val.1;

        Ok(Filter::Expr(name, op, val.0))
    }

    // This tokenizer has two states:
    //    hasexpr (allows And, Or, Close)
    //   !hasexpr (allows Open, Expr)
    // These states are necessary to handle ambiguity between Expr and the And/Or tokens, and are
    // also used to enforce the following properties (which simplifies the parsing step):
    // - Expr and And/Or tokens cannot be chained
    // - And/Or/Close tokens always follow a Close/Expr token.
    // - Expr/Open tokens always follow a Open/And/Or token
    //
    // An Expr token doesn't consume anything, the caller is expected to run token_expr() to get
    // the expression and advance the parsing state.
    fn token(&mut self) -> Result<Token> {
        self.conws();

        let ret = match (self.hasexpr, self.buf.chars().next()) {
            (_,    None)      => Err("Unexpected end of input"),
            (false,Some('(')) => { self.con(1); Ok(Token::Open) },
            (true, Some(')')) => { self.con(1); Ok(Token::Close) },
            (true, Some('a')) => if self.buf.starts_with("and") { self.con(3); Ok(Token::And) } else { Err("Invalid token") },
            (true, Some('o')) => if self.buf.starts_with("or")  { self.con(2); Ok(Token::Or ) } else { Err("Invalid token") },
            (false,_)         => Ok(Token::Expr),
            _ => Err("Invalid token"),
        };

        self.hasexpr = match ret { Ok(Token::Close) | Ok(Token::Expr) => true, _ => false };
        ret
    }

    fn parse(&mut self) -> Result<Filter> {
        // This is a simple shunting-yard implementation
        let mut exp = Vec::new();
        let mut ops = vec![Token::Open]; // Only And, Or and Open

        if self.token()? != Token::Open {
            return Err("Filter must start with an open parentheses");
        }

        while ops.len() > 0 {
            match self.token()? {
                Token::Expr => exp.push(self.token_expr()?),

                op@Token::Open => ops.push(op),

                Token::Close => {
                    while let Some(op) = ops.pop() {
                        if op == Token::Open {
                            break;
                        } else {
                            apply(&mut exp, op);
                        }
                    }
                },

                o1@Token::And | o1@Token::Or => {
                    while let Some(&o2) = ops.last() {
                        if o2 != Token::Open && (o1 != o2 && o1 == Token::Or) {
                            ops.pop();
                            apply(&mut exp, o2);
                        } else {
                            break;
                        }
                    }
                    ops.push(o1);
                },
            }
        }
        Ok(exp.pop().unwrap())
    }
}


fn apply(exp: &mut Vec<Filter>, op: Token) {
    let right = Box::new(exp.pop().unwrap());
    let left  = Box::new(exp.pop().unwrap());
    exp.push(if op == Token::And { Filter::And(left, right) } else { Filter::Or(left, right) });
}


fn parse_op(s: &str) -> Option<Op> {
    if      s.starts_with("=" ) { Some(Op::Eq ) }
    else if s.starts_with("!=") { Some(Op::NEq) }
    else if s.starts_with("<=") { Some(Op::LEq) }
    else if s.starts_with("<" ) { Some(Op::Le ) }
    else if s.starts_with(">=") { Some(Op::GEq) }
    else if s.starts_with(">" ) { Some(Op::Gt ) }
    else if s.starts_with("~" ) { Some(Op::Fuzzy) }
    else { None }
}


pub fn parse_filter(s: &str) -> Result<(Filter, &str)> {
    let mut p = FilterParser{buf: s, hasexpr: false};
    p.parse().map(|r| (r, p.buf))
}


pub fn parse_arg(s: &str) -> Result<(Arg, &str)> {
    let s = trim_ws(s);

    // This match on the first character can be replaced by simply trying parse_filter and
    // parse_json in sequence; but that results in less than ideal error messages on badly
    // formatted input.
    match s.chars().next() {
        None => return Err("Empty argument"),

        Some('(') => {
            return parse_filter(s).map(|(v,r)| (Arg::Filter(v), r));
        },

        Some('[') | Some('{') | Some('"') => {
            return parse_json(s).map(|(v,r)| (Arg::Json(v), r));
        },

        Some(_) => {
            if let Ok((v,r)) = parse_json(s) {
                return Ok((Arg::Json(v), r));

            } else {
                let mut splt = s.splitn(2, is_ws);
                let v = splt.next().unwrap();
                let rem = splt.next().unwrap_or("");

                if !v.contains(|c| !is_barestr(c)) {
                    return Ok((Arg::BareString(v.to_string()), rem));
                } else {
                    return Err("Invalid argument")
                }
            }
        },
    }
}


pub fn parse_message(s: &str) -> Result<Message> {
    let mut buf = trim_ws(s);

    let mut splt = buf.splitn(2, is_ws);
    let name = splt.next().ok_or("Empty message")?;
    let mut msg = Message::new(name)?;

    buf = trim_ws(splt.next().unwrap_or(""));
    while buf.len() > 0 {
        let v = parse_arg(buf)?;
        msg = msg.push_arg(v.0);
        buf = trim_ws(v.1);
    }

    Ok(msg)
}


#[test]
fn test_parse_filter() {
    let ok = |i, o| {
        let s = format!("{}garbage", i);
        let f = parse_filter(&s).unwrap();
        assert_eq!(&format!("{}", f.0), o);
        assert_eq!(f.1, "garbage");
    };
    ok("(n=1)", "(n = 1)");
    ok("(something_else>=[1,\"str\"])", "(something_else >= [1,\"str\"])");
    ok("(((n=1) and blah=[]))", "((n = 1) and (blah = []))");
    ok("(((n=1) and blah=[] or x=\"hi\"))", "(((n = 1) and (blah = [])) or (x = \"hi\"))");
    ok("(a=1andb=2andc=3)", "((a = 1) and ((b = 2) and (c = 3)))");
    ok("(a=1orb=2andc=3)", "((a = 1) or ((b = 2) and (c = 3)))");
    ok("(a=1orb=2andc=3and(d=4ande=5orf=6)andg=7)", "((a = 1) or ((b = 2) and ((c = 3) and ((((d = 4) and (e = 5)) or (f = 6)) and (g = 7)))))");
    ok("(and=nulloror!=false)", "((and = null) or (or != false))");

    let nok = |i| { assert!(parse_filter(i).is_err()) };
    nok("()");
    nok("(n=1 n=1)");
    nok("n=1");
    nok("(and)");
    nok("(n=1 and");
    nok("(n=1 and )");
    nok("(n=1 and and n=2)");
    nok(") and n=1");
}