Class: Fusion::Parser

Inherits:
Object
  • Object
show all
Includes:
AST
Defined in:
lib/fusion/parser.rb

Constant Summary collapse

PRIMARY_STARTERS =

Tokens that can begin a primary expression (used by parse_prefix to decide whether ‘!` is followed by an operand).

%i[number string true_kw false_kw null_kw bang
lbracket lbrace lparen ident at].freeze
GUARDEDPAT_STARTERS =

Tokens that can begin a ‘guardedpat` (used to detect whether `!` is followed by a payload pattern or stands alone).

%i[number string true_kw false_kw null_kw
lbracket lbrace ident].freeze

Constants included from AST

AST::ArrayItem, AST::ArraySpread, AST::Clause, AST::Identifier, AST::KeyValuePair, AST::ObjectSpread, AST::PatternItem, AST::PatternPair, AST::PatternRest

Class Method Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(tokens) ⇒ Parser

Returns a new instance of Parser.



18
19
20
21
 
# File 'lib/fusion/parser.rb', line 18

def initialize(tokens)
  @toks = tokens
  @i = 0
end

Class Method Details

.parse_file(src, location:) ⇒ Object

Parse a complete program. The lexer and parser report failures by raising ParseError; this single entry point rescues them and returns a standardized syntax_error value, so no caller ever sees a raw Ruby error. ‘location` is the syntax_error’s “code X” / “code <inline>” context.



27
28
29
30
31
32
33
34
35
 
# File 'lib/fusion/parser.rb', line 27

def self.parse_file(src, location:)
  toks = Lexer.new(src).tokens
  p = new(toks)
  expr = p.parse_expr
  p.expect(:eof)
  expr
rescue ParseError => err
  Interpreter::ErrorVal.internal(kind: "syntax_error", location: location, operation: "parsing", input: src, message: err.message)
end

.parse_repl(src, location:) ⇒ Object

Parse one REPL entry — a statement (‘identifier “=” expr`) or a bare expression — returning an AST::Statement::Assignment / AST::Expression, or, like parse_file, a standardized syntax_error value instead of ever raising. The REPL uses the error/non-error distinction to tell “keep editing” (didn’t parse yet) from “evaluate now” (a complete statement or expression).



42
43
44
45
46
47
48
49
50
 
# File 'lib/fusion/parser.rb', line 42

def self.parse_repl(src, location:)
  toks = Lexer.new(src).tokens
  p = new(toks)
  entry = p.parse_repl_entry
  p.expect(:eof)
  entry
rescue ParseError => err
  Interpreter::ErrorVal.internal(kind: "syntax_error", location: location, operation: "parsing", input: src, message: err.message)
end

Instance Method Details

#advance ⇒ Object 



397
 
# File 'lib/fusion/parser.rb', line 397

def advance = (@toks[@i].tap { @i += 1 })

#at?(type) ⇒ Boolean

Returns:

  • (Boolean) 


396
 
# File 'lib/fusion/parser.rb', line 396

def at?(type) = peek.type == type

#expect(type) ⇒ Object

Raises:



398
399
400
401
402
 
# File 'lib/fusion/parser.rb', line 398

def expect(type)
  t = peek
  raise ParseError, "Expected #{type} but got #{t.type} (#{t.value.inspect}) at #{t.pos}" unless t.type == type
  advance
end

#looks_like_function? ⇒ Boolean

Look ahead from current position (just after “(”) to decide if this is a function literal: is there a top-level “=>” before the matching “)”?

Returns:

  • (Boolean) 


242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
 
# File 'lib/fusion/parser.rb', line 242

def looks_like_function?
  depth = 0
  j = @i
  while j < @toks.length
    t = @toks[j]
    case t.type
    when :lparen, :lbracket, :lbrace then depth += 1
    when :rparen, :rbracket, :rbrace
      return false if depth.zero? # hit our closing ) first
      depth -= 1
    when :arrow
      return true if depth.zero?
    when :eof
      return false
    end
    j += 1
  end
  false
end

#parse_array ⇒ Object 



164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
 
# File 'lib/fusion/parser.rb', line 164

def parse_array
  expect(:lbracket)
  items = []
  until at?(:rbracket)
    if at?(:spread)
      advance
      items << ArraySpread.new(value: parse_expr)
    else
      items << ArrayItem.new(value: parse_expr)
    end
    break unless at?(:comma)
    advance
  end
  expect(:rbracket)
  Expression::ArrLit.new(items: items)
end

#parse_arraypat ⇒ Object

p_array (reference.md §2.5). Items are ‘p_guarded`s — never error patterns. The grammar’s two arms (with / without a rest) become two phases: the loop parses leading items up to an optional single ‘…rest`; once a rest is consumed, the inner loop parses trailing items only, so a second `…` lands in `parse_guardedpat` as an unexpected token. There is no `seen_rest` flag —“at most one rest” is enforced by the shape of the loop.



328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
 
# File 'lib/fusion/parser.rb', line 328

def parse_arraypat
  expect(:lbracket)
  items = []
  until at?(:rbracket)
    if at?(:spread)
      items << parse_pattern_rest
      while at?(:comma)
        advance
        break if at?(:rbracket) # trailing comma
        raise ParseError, "a pattern may contain at most one `...rest` (at #{peek.pos})" if at?(:spread)
        items << PatternItem.new(pattern: parse_guardedpat)
      end
      break
    end
    items << PatternItem.new(pattern: parse_guardedpat)
    break unless at?(:comma)
    advance
  end
  expect(:rbracket)
  Pattern::PArr.new(items: items)
end

#parse_corepat ⇒ Object 



303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
 
# File 'lib/fusion/parser.rb', line 303

def parse_corepat
  t = peek
  case t.type
  when :number, :string then advance; Pattern::PLit.new(value: t.value)
  when :true_kw, :false_kw, :null_kw then advance; Pattern::PLit.new(value: t.value)
  when :lbracket then parse_arraypat
  when :lbrace then parse_objectpat
  when :ident
    advance
    t.value == "_" ? Pattern::PWild.new(dummy: nil) : Pattern::PBind.new(name: t.value)
  when :bang
    # `!pat` is only valid as a clause's top-level pattern, never inside an
    # array element, object member, or error payload.
    raise ParseError, "`!pat` may only appear as a clause's top-level pattern (at #{t.pos})"
  else
    raise ParseError, "Unexpected token in pattern: #{t.type} at #{t.pos}"
  end
end

#parse_errpat ⇒ Object 



281
282
283
284
285
286
287
288
 
# File 'lib/fusion/parser.rb', line 281

def parse_errpat
  expect(:bang)
  if GUARDEDPAT_STARTERS.include?(peek.type)
    Pattern::PErr.new(inner: parse_guardedpat)               # "!" guardedpat
  else
    Pattern::PErr.new(inner: Pattern::PWild.new(dummy: nil)) # bare "!" — matches any error, binds nothing
  end
end

#parse_expr ⇒ Object 



69
70
71
 
# File 'lib/fusion/parser.rb', line 69

def parse_expr
  parse_pipe
end

#parse_fileref ⇒ Object 



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
 
# File 'lib/fusion/parser.rb', line 137

def parse_fileref
  expect(:at)
  # Bare "@" = current file: not followed by something that can begin a path.
  nxt = peek
  starts_path = (nxt.type == :ident) || (nxt.type == :dot && peek(1)&.type == :dot)
  return Expression::FileRef.new(variety: :self, path: nil) unless starts_path
  # refpath: { "../" } segment { "/" segment }
  parts = []
  has_dotdot = false
  while at?(:dot) && peek(1)&.type == :dot
    advance; advance # consume the two dots of ..
    parts << ".."
    expect(:slash)
    has_dotdot = true
  end
  parts << expect(:ident).value
  while at?(:slash)
    advance
    parts << expect(:ident).value
  end
  # A reference is eligible for builtin/stdlib fallback (:name) iff it does NOT
  # contain "../". Downward paths like "dir/a" are still eligible; only "../"
  # (escaping upward) forces pure file-path (:path) resolution.
  bare = !has_dotdot
  Expression::FileRef.new(variety: bare ? :name : :path, path: parts.join("/"))
end

#parse_function_or_group ⇒ Object

A “(” begins a grouped expression, a function literal, or — when empty —the clause-less function ‘()`. A function is a comma-separated list of `pattern => expr`; we detect one by scanning for a top-level `=>` before the matching `)`. `()` matches nothing (so it yields null for any normal input and propagates errors).



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
 
# File 'lib/fusion/parser.rb', line 211

def parse_function_or_group
  expect(:lparen)
  if at?(:rparen)
    advance
    return Expression::FuncLit.new(clauses: [])
  end
  if looks_like_function?
    clauses = []
    loop do
      pat = parse_pattern
      expect(:arrow)
      body = parse_expr
      clauses << Clause.new(pattern: pat, body: body)
      if at?(:comma)
        advance
        break if at?(:rparen) # trailing comma
      else
        break
      end
    end
    expect(:rparen)
    Expression::FuncLit.new(clauses: clauses)
  else
    e = parse_expr
    expect(:rparen)
    e
  end
end

#parse_guardedpat ⇒ Object 



290
291
292
293
294
295
296
297
298
299
300
301
 
# File 'lib/fusion/parser.rb', line 290

def parse_guardedpat
  inner = parse_corepat
  if at?(:question)
    advance
    # A predicate is a full pipe so it may chain functions: `a ? b | c` tests
    # `a | b | c`. It stops at `=>`, `,`, `]`, `}`, `)` like any expression.
    pred = parse_pipe
    Pattern::PGuard.new(inner: inner, pred_expr: pred)
  else
    inner
  end
end

#parse_object ⇒ Object

Fixed keys must be distinct (the ObjLit data rule); a repeat is a clean syntax_error. Keys arriving via ‘…spread` are dynamic and not checked.



183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
 
# File 'lib/fusion/parser.rb', line 183

def parse_object
  expect(:lbrace)
  pairs = []
  keys = []
  until at?(:rbrace)
    if at?(:spread)
      advance
      pairs << ObjectSpread.new(value: parse_expr)
    else
      key_tok = expect(:string)
      key = key_tok.value
      raise ParseError, "duplicate key #{key.inspect} (at #{key_tok.pos})" if keys.include?(key)
      keys << key
      expect(:colon)
      pairs << KeyValuePair.new(key: key, value: parse_expr)
    end
    break unless at?(:comma)
    advance
  end
  expect(:rbrace)
  Expression::ObjLit.new(pairs: pairs)
end

#parse_objectpat ⇒ Object

p_object (reference.md §2.5). Leading pairs up to an optional single ‘…rest`, which must come last — only a trailing comma may follow it. Keys must be distinct (the PObj data rule); a repeat is a clean syntax_error.



353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
 
# File 'lib/fusion/parser.rb', line 353

def parse_objectpat
  expect(:lbrace)
  pairs = []
  keys = []
  until at?(:rbrace)
    if at?(:spread)
      pairs << parse_pattern_rest
      advance if at?(:comma) && peek(1)&.type == :rbrace # trailing comma
      unless at?(:rbrace)
        raise ParseError, "in an object pattern, `...rest` must come last (at #{peek.pos})"
      end
      break
    end
    key_pos = peek.pos
    pair = parse_pattern_pair
    raise ParseError, "duplicate key #{pair.key.inspect} (at #{key_pos})" if keys.include?(pair.key)
    keys << pair.key
    pairs << pair
    break unless at?(:comma)
    advance
  end
  expect(:rbrace)
  Pattern::PObj.new(pairs: pairs)
end

#parse_pattern ⇒ Object

—- Patterns —- —- Pattern grammar (mirrors reference.md §2.5 EBNF) ——————

pattern   = p_error | p_guarded
p_error   = "!" | "!" p_guarded
p_guarded = p_core [ "?" predicate ]
p_core    = p_literal | p_bind | p_wildcard | p_array | p_object

Note: ‘p_core` does NOT include p_error. The “no nested !pat” property falls out of the grammar shape — `p_error` is only reachable from `pattern` (a clause’s top level), never from inside arrays, objects, or another error’s payload. No flag-threading is needed.



272
273
274
 
# File 'lib/fusion/parser.rb', line 272

def parse_pattern
  at?(:bang) ? parse_errpat : parse_guardedpat
end

#parse_pattern_pair ⇒ Object

p_pair = string “:” p_guarded



388
389
390
391
392
 
# File 'lib/fusion/parser.rb', line 388

def parse_pattern_pair
  key = expect(:string).value
  expect(:colon)
  PatternPair.new(key: key, pattern: parse_guardedpat)
end

#parse_pattern_rest ⇒ Object

p_rest = “…” [ identifier ] — the single rest binder, shared by array and object patterns. Callers parse it only at a rest position and then continue with items/pairs only, which is what holds a pattern to one rest.



381
382
383
384
385
 
# File 'lib/fusion/parser.rb', line 381

def parse_pattern_rest
  expect(:spread)
  name = at?(:ident) ? advance.value : nil
  PatternRest.new(name: name)
end

#parse_pipe ⇒ Object 



73
74
75
76
77
78
79
80
81
 
# File 'lib/fusion/parser.rb', line 73

def parse_pipe
  left = parse_prefix
  while at?(:pipe)
    advance
    right = parse_prefix
    left = Expression::Pipe.new(left: left, right: right)
  end
  left
end

#parse_postfix ⇒ Object 



104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
 
# File 'lib/fusion/parser.rb', line 104

def parse_postfix
  node = parse_primary
  loop do
    if at?(:dot)
      advance
      key = expect(:ident).value
      node = Expression::Member.new(obj: node, key: key)
    elsif at?(:lbracket)
      advance
      idx = parse_expr
      expect(:rbracket)
      node = Expression::Index.new(obj: node, idx: idx)
    else
      break
    end
  end
  node
end

#parse_prefix ⇒ Object

‘!` is a prefix operator that constructs an error from its operand. A bare `!` (no operand follows) is shorthand for `!null`. Binds tighter than `|` so `!x | f` is `(!x) | f`; looser than postfix so `!x.foo` is `!(x.foo)`.



91
92
93
94
95
96
97
98
99
100
101
102
 
# File 'lib/fusion/parser.rb', line 91

def parse_prefix
  if at?(:bang)
    advance
    if PRIMARY_STARTERS.include?(peek.type)
      Expression::ErrLit.new(payload: parse_prefix)   # allow !!x to nest
    else
      Expression::ErrLit.new(payload: nil)            # bare ! -> !null
    end
  else
    parse_postfix
  end
end

#parse_primary ⇒ Object 



123
124
125
126
127
128
129
130
131
132
133
134
135
 
# File 'lib/fusion/parser.rb', line 123

def parse_primary
  t = peek
  case t.type
  when :number, :string then advance; Expression::Lit.new(value: t.value)
  when :true_kw, :false_kw, :null_kw then advance; Expression::Lit.new(value: t.value)
  when :lbracket then parse_array
  when :lbrace then parse_object
  when :lparen then parse_function_or_group
  when :ident then advance; Expression::Ident.new(name: t.value)
  when :at then parse_fileref
  else raise ParseError, "Unexpected token #{t.type} (#{t.value.inspect}) at #{t.pos}"
  end
end

#parse_repl_entry ⇒ Object

A leading ‘identifier =` marks a statement; anything else is an expression. (A bare identifier is itself a valid expression, so the `=` is the decider.)



54
55
56
57
58
59
60
 
# File 'lib/fusion/parser.rb', line 54

def parse_repl_entry
  if at?(:ident) && peek(1)&.type == :equals
    parse_statement
  else
    parse_expr
  end
end

#parse_statement ⇒ Object

statement = identifier “=” expr (REPL only; files contain one expr)



63
64
65
66
67
 
# File 'lib/fusion/parser.rb', line 63

def parse_statement
  name = expect(:ident).value
  expect(:equals)
  AST::Statement::Assignment.new(name: name, expression: parse_expr)
end

#peek(o = 0) ⇒ Object

—- token helpers —-



395
 
# File 'lib/fusion/parser.rb', line 395

def peek(o = 0) = @toks[@i + o]