mir-analyzer 0.55.0

Analysis engine for the mir PHP static analyzer
Documentation
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
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
use mir_types::{ArrayKey, Atomic, Name, Type};
use php_ast::ast::BinaryOp;
use php_ast::owned::{Expr, ExprKind};
use rustc_hash::FxHashSet;

use crate::subtype::is_subtype;

/// PHP canonicalizes a numeric string array key (e.g. `"0"`, `"42"`, `"-5"`)
/// to an int key at runtime — `$arr['0']` and `$arr[0]` are the same slot.
/// Returns the canonical int when `s` is such a string; `None` means `s`
/// stays a string key (e.g. `"01"`, `"+1"`, `"-0"`, `"1.0"`, `""`).
pub fn canonical_int_array_key(s: &str) -> Option<i64> {
    let bytes = s.as_bytes();
    if bytes.is_empty() {
        return None;
    }
    let (neg, digits) = if bytes[0] == b'-' {
        (true, &bytes[1..])
    } else {
        (false, bytes)
    };
    if digits.is_empty() || !digits.iter().all(u8::is_ascii_digit) {
        return None;
    }
    // No leading zero unless the value is exactly "0"; PHP also treats "-0"
    // as a non-canonical string key.
    if digits.len() > 1 && digits[0] == b'0' {
        return None;
    }
    if neg && digits == b"0" {
        return None;
    }
    std::str::from_utf8(digits)
        .ok()?
        .parse::<i64>()
        .ok()
        .map(|v| if neg { -v } else { v })
}

/// Resolve an index expression to a literal array key, canonicalizing numeric
/// string keys (`"0"` → `0`) the same way [`canonical_int_array_key`] does.
/// Returns `None` for a dynamic (non-literal) index.
pub fn literal_array_key_of_kind(kind: &ExprKind) -> Option<ArrayKey> {
    match kind {
        ExprKind::String(s) => Some(match canonical_int_array_key(s) {
            Some(i) => ArrayKey::Int(i),
            None => ArrayKey::String(std::sync::Arc::from(s.as_ref())),
        }),
        ExprKind::Int(i) => Some(ArrayKey::Int(*i)),
        _ => None,
    }
}

/// Update a nested shape write (`$arr['a']['b'] = $v`) by walking into the
/// matching per-key property at each level instead of widening the whole
/// outer shape into a generic array. `path` is ordered innermost-first (the
/// key directly on `current`, then progressively outer keys) and `leaf_value`
/// is the value being assigned at the final (outermost) key.
///
/// Returns `None` when the shape at any level doesn't cleanly resolve (an
/// unknown key, a non-uniform union, a non-shape atom, …) so the caller can
/// fall back to the existing generic accumulator.
pub fn set_nested_keyed_value(
    current: &Type,
    path: &[ArrayKey],
    leaf_value: &Type,
) -> Option<Type> {
    let (key, rest) = path.split_first()?;
    if current.types.is_empty() {
        return None;
    }
    let all_shapes_have_key = current.types.iter().all(
        |a| matches!(a, Atomic::TKeyedArray { properties, .. } if properties.contains_key(key)),
    );
    if !all_shapes_have_key {
        return None;
    }
    let mut result = Type::empty();
    result.possibly_undefined = current.possibly_undefined;
    result.from_docblock = current.from_docblock;
    for atomic in &current.types {
        let Atomic::TKeyedArray {
            properties,
            is_open,
            is_list,
        } = atomic
        else {
            unreachable!("filtered to TKeyedArray above")
        };
        let mut new_properties = properties.clone();
        let existing = properties.get(key).expect("checked by all_shapes_have_key");
        let new_inner = if rest.is_empty() {
            leaf_value.clone()
        } else {
            set_nested_keyed_value(&existing.ty, rest, leaf_value)?
        };
        new_properties.insert(
            key.clone(),
            mir_types::atomic::KeyedProperty {
                ty: new_inner,
                optional: false,
            },
        );
        result.add_type(Atomic::TKeyedArray {
            properties: new_properties,
            is_open: *is_open,
            is_list: *is_list,
        });
    }
    Some(result)
}

/// Remove `key` from every `TKeyedArray` atomic in `ty`'s union that has it,
/// leaving all other atoms and properties unchanged. Used for
/// `unset($arr['key'])`, which genuinely removes the key from the array
/// (regardless of whether the shape is open or closed) rather than merely
/// marking it optional.
pub fn remove_key_from_shapes(ty: &Type, key: &ArrayKey) -> Type {
    let mut changed = false;
    let mut result = Type::empty();
    for a in &ty.types {
        if let Atomic::TKeyedArray {
            properties,
            is_open,
            is_list,
        } = a
        {
            if properties.contains_key(key) {
                changed = true;
                let mut new_props = properties.clone();
                new_props.shift_remove(key);
                result.add_type(Atomic::TKeyedArray {
                    properties: new_props,
                    is_open: *is_open,
                    is_list: *is_list,
                });
                continue;
            }
        }
        result.add_type(a.clone());
    }
    if !changed {
        return ty.clone();
    }
    result.from_docblock = ty.from_docblock;
    result
}

pub fn widen_array_with_value_and_key(
    current: &Type,
    new_value: &Type,
    new_key: &Type,
    literal_key: Option<&mir_types::ArrayKey>,
) -> Type {
    // Overwriting an EXISTING literal key on a shape (`$arr['a'] = 2;` where
    // 'a' is already a known property) updates just that one property,
    // leaving every other key's type untouched — routing this through the
    // generic accumulator below would collapse the whole shape into a wide
    // `array<K, V>` union even though no other key was affected by the write.
    if let Some(key) = literal_key {
        let all_shapes_have_key = !current.types.is_empty()
            && current.types.iter().all(|a| match a {
                Atomic::TKeyedArray { properties, .. } => properties.contains_key(key),
                _ => false,
            });
        if all_shapes_have_key {
            let mut result = Type::empty();
            result.possibly_undefined = current.possibly_undefined;
            result.from_docblock = current.from_docblock;
            for atomic in &current.types {
                let Atomic::TKeyedArray {
                    properties,
                    is_open,
                    is_list,
                } = atomic
                else {
                    unreachable!("filtered to TKeyedArray above")
                };
                let mut new_properties = properties.clone();
                // The key is now definitely assigned on this path, regardless
                // of whether it was previously optional.
                new_properties.insert(
                    key.clone(),
                    mir_types::atomic::KeyedProperty {
                        ty: new_value.clone(),
                        optional: false,
                    },
                );
                result.add_type(Atomic::TKeyedArray {
                    properties: new_properties,
                    is_open: *is_open,
                    is_list: *is_list,
                });
            }
            return result;
        }
    }

    let mut result = Type::empty();
    result.possibly_undefined = current.possibly_undefined;
    result.from_docblock = current.from_docblock;
    let mut found_array = false;
    // Merge ALL array-like variants from current into a single accumulated TArray/TList.
    // Without this, each TArray variant in a growing union independently emits a new TArray,
    // causing unbounded union growth across salsa fixpoint iterations (infinite recursion).
    let mut acc_key: Option<Type> = None;
    let mut acc_value: Option<Type> = None;
    let mut acc_list: Option<Type> = None;
    for atomic in &current.types {
        match atomic {
            Atomic::TKeyedArray { properties, .. } => {
                let mut all_values = new_value.clone();
                let mut all_keys = new_key.clone();
                for prop in properties.values() {
                    all_values.merge_with(&prop.ty);
                }
                for k in properties.keys() {
                    let key_atomic = match k {
                        mir_types::ArrayKey::String(s) => Atomic::TLiteralString(s.clone()),
                        mir_types::ArrayKey::Int(i) => Atomic::TLiteralInt(*i),
                    };
                    all_keys.merge_with(&Type::single(key_atomic));
                }
                fold_into(&mut acc_key, all_keys);
                fold_into(&mut acc_value, all_values);
                found_array = true;
            }
            Atomic::TArray { key, value } => {
                fold_into(&mut acc_key, Type::merge(key, new_key));
                fold_into(&mut acc_value, Type::merge(value, new_value));
                found_array = true;
            }
            Atomic::TList { value } | Atomic::TNonEmptyList { value } => {
                fold_into(&mut acc_list, Type::merge(value, new_value));
                found_array = true;
            }
            Atomic::TNonEmptyArray { key, value } => {
                fold_into(&mut acc_key, Type::merge(key, new_key));
                fold_into(&mut acc_value, Type::merge(value, new_value));
                found_array = true;
            }
            Atomic::TMixed => {
                return Type::mixed();
            }
            other => {
                result.add_type(other.clone());
            }
        }
    }
    if let (Some(key), Some(value)) = (acc_key, acc_value) {
        result.add_type(Atomic::TArray {
            key: Box::new(key),
            value: Box::new(value),
        });
    }
    if let Some(v) = acc_list {
        result.add_type(Atomic::TList { value: Box::new(v) });
    }
    if !found_array {
        return current.clone();
    }
    result
}

/// Widen an existing array-like type by appending `new_value` via push notation (`[]`).
/// Always produces `TList { merged_value }`, regardless of the current key type,
/// because push notation in PHP assigns the next integer index.
pub fn widen_array_as_list(current: &Type, new_value: &Type) -> Type {
    let mut result = Type::empty();
    result.possibly_undefined = current.possibly_undefined;
    result.from_docblock = current.from_docblock;
    let mut acc: Option<Type> = Some(new_value.clone());
    let mut found_array = false;
    for atomic in &current.types {
        match atomic {
            Atomic::TKeyedArray { properties, .. } => {
                for prop in properties.values() {
                    fold_into(&mut acc, prop.ty.clone());
                }
                found_array = true;
            }
            Atomic::TArray { value, .. }
            | Atomic::TNonEmptyArray { value, .. }
            | Atomic::TList { value }
            | Atomic::TNonEmptyList { value } => {
                fold_into(&mut acc, *value.clone());
                found_array = true;
            }
            Atomic::TMixed => return Type::mixed(),
            other => result.add_type(other.clone()),
        }
    }
    if !found_array {
        return current.clone();
    }
    if let Some(v) = acc {
        result.add_type(Atomic::TList { value: Box::new(v) });
    }
    result
}

fn fold_into(acc: &mut Option<Type>, new: Type) {
    match acc {
        None => *acc = Some(new),
        Some(existing) => existing.merge_with(&new),
    }
}

/// The inclusive integer bounds of `ty` when it is an integer-only type, as
/// `(min, max)` where `None` means unbounded on that side. Returns `None` when
/// any member is not an integer (so the caller falls back to scalar inference).
/// Literals are exact bounds; a general `int` is unbounded both ways.
fn int_bounds(ty: &Type) -> Option<(Option<i64>, Option<i64>)> {
    if ty.types.is_empty() {
        return None;
    }
    let mut min: Option<i64> = Some(i64::MAX);
    let mut max: Option<i64> = Some(i64::MIN);
    for a in &ty.types {
        let (lo, hi) = match a {
            Atomic::TLiteralInt(n) => (Some(*n), Some(*n)),
            Atomic::TIntRange { min, max } => (*min, *max),
            // Named int subtypes carry implicit bounds: use them so arithmetic
            // like `positive-int + 1` yields `int<2, max>` rather than bare `int`.
            Atomic::TPositiveInt => (Some(1), None),
            Atomic::TNonNegativeInt => (Some(0), None),
            Atomic::TNegativeInt => (None, Some(-1)),
            Atomic::TInt => (None, None),
            _ => return None,
        };
        // Widen the accumulated bounds to cover this member (union semantics).
        min = match (min, lo) {
            (Some(m), Some(l)) => Some(m.min(l)),
            _ => None,
        };
        max = match (max, hi) {
            (Some(m), Some(h)) => Some(m.max(h)),
            _ => None,
        };
    }
    Some((min, max))
}

/// Whether `ty` carries an explicit integer range or a named int subtype with
/// known implicit bounds (positive-int, non-negative-int, negative-int).
fn contains_int_range(ty: &Type) -> bool {
    ty.types.iter().any(|a| {
        matches!(
            a,
            Atomic::TIntRange { .. }
                | Atomic::TPositiveInt
                | Atomic::TNonNegativeInt
                | Atomic::TNegativeInt
        )
    })
}

/// Range-aware integer arithmetic for `+` and `-`: when at least one operand is
/// an integer range (e.g. a `count()` result), propagate faithful bounds so
/// `count($a) + 1` is `int<1, max>` and `count($a) - 1` is `int<-1, max>`.
/// Returns `None` for anything else (including literal-only arithmetic, left to
/// [`infer_arithmetic`] so it is not perturbed).
fn as_single_literal_int(ty: &Type) -> Option<i64> {
    if ty.types.len() == 1 {
        if let Atomic::TLiteralInt(n) = &ty.types[0] {
            return Some(*n);
        }
    }
    None
}

pub fn infer_int_range_arithmetic(left: &Type, right: &Type, op: BinaryOp) -> Option<Type> {
    // Fast path: both operands are known literal ints — fold at analysis time.
    if let (Some(l), Some(r)) = (as_single_literal_int(left), as_single_literal_int(right)) {
        let result = match op {
            BinaryOp::Add => l.checked_add(r),
            BinaryOp::Sub => l.checked_sub(r),
            BinaryOp::Mul => l.checked_mul(r),
            // Integer division only when divisor is nonzero and result is exact.
            BinaryOp::Div if r != 0 && l % r == 0 => Some(l / r),
            BinaryOp::Mod if r != 0 => Some(l % r),
            _ => None,
        };
        if let Some(n) = result {
            return Some(Type::single(Atomic::TLiteralInt(n)));
        }
        // Non-exact literal integer division (e.g. 5 / 2 = 2.5) → float.
        if op == BinaryOp::Div && r != 0 {
            return Some(Type::single(Atomic::TFloat));
        }
    }

    // Only engage when a genuine range is in play; plain int/literal operands
    // keep the existing scalar inference.
    if !contains_int_range(left) && !contains_int_range(right) {
        return None;
    }
    let (lmin, lmax) = int_bounds(left)?;
    let (rmin, rmax) = int_bounds(right)?;
    let add = |a: Option<i64>, b: Option<i64>| match (a, b) {
        (Some(a), Some(b)) => a.checked_add(b),
        _ => None,
    };
    let sub = |a: Option<i64>, b: Option<i64>| match (a, b) {
        (Some(a), Some(b)) => a.checked_sub(b),
        _ => None,
    };
    let mul_opt = |a: Option<i64>, b: Option<i64>| match (a, b) {
        (Some(a), Some(b)) => a.checked_mul(b),
        _ => None,
    };
    let (min, max) = match op {
        BinaryOp::Add => (add(lmin, rmin), add(lmax, rmax)),
        // [lmin,lmax] - [rmin,rmax] = [lmin - rmax, lmax - rmin]
        BinaryOp::Sub => (sub(lmin, rmax), sub(lmax, rmin)),
        // Multiplication: only handle the case where both operands are non-negative,
        // which is the common case (`count * stride`, `width * height`, etc.).
        // lmin/rmin must be Some(>=0) — None means unbounded below, i.e., can be negative.
        // For mixed-sign operands the four-corner product is complex; defer to infer_arithmetic.
        BinaryOp::Mul if lmin.is_some_and(|m| m >= 0) && rmin.is_some_and(|m| m >= 0) => {
            (mul_opt(lmin, rmin), mul_opt(lmax, rmax))
        }
        // Modulo: result range depends only on the divisor.
        // For a known positive divisor K: result ∈ [0, K-1] when dividend ≥ 0,
        // or [-(K-1), K-1] when dividend may be negative (PHP truncates toward zero).
        BinaryOp::Mod if rmin == rmax && rmin.is_some_and(|r| r > 0) => {
            let divisor = rmin.unwrap();
            if lmin.is_some_and(|m| m >= 0) {
                (Some(0), Some(divisor - 1))
            } else {
                (Some(-(divisor - 1)), Some(divisor - 1))
            }
        }
        _ => return None,
    };
    Some(Type::single(Atomic::TIntRange { min, max }))
}

/// Bool and null coerce to int (0/1 and 0 respectively) in PHP arithmetic;
/// they never produce float. This predicate is used in `infer_arithmetic` to
/// extend the "returns int" condition beyond pure-int operands.
fn coerces_to_int_in_arithmetic(t: &Atomic) -> bool {
    t.is_int()
        || matches!(
            t,
            Atomic::TBool | Atomic::TTrue | Atomic::TFalse | Atomic::TNull
        )
}

pub fn infer_arithmetic(left: &Type, right: &Type) -> Type {
    if left.is_mixed() || right.is_mixed() {
        return Type::mixed();
    }

    let left_is_array = left.contains(|t| {
        matches!(
            t,
            Atomic::TArray { .. }
                | Atomic::TNonEmptyArray { .. }
                | Atomic::TList { .. }
                | Atomic::TNonEmptyList { .. }
                | Atomic::TKeyedArray { .. }
        )
    });
    let right_is_array = right.contains(|t| {
        matches!(
            t,
            Atomic::TArray { .. }
                | Atomic::TNonEmptyArray { .. }
                | Atomic::TList { .. }
                | Atomic::TNonEmptyList { .. }
                | Atomic::TKeyedArray { .. }
        )
    });
    if left_is_array || right_is_array {
        let merged_left = if left_is_array {
            left.clone()
        } else {
            Type::single(Atomic::TArray {
                key: Box::new(Type::single(Atomic::TMixed)),
                value: Box::new(Type::mixed()),
            })
        };
        return merged_left;
    }

    let left_is_float = left.contains(|t| {
        matches!(
            t,
            Atomic::TFloat | Atomic::TIntegralFloat | Atomic::TLiteralFloat(..)
        )
    });
    let right_is_float = right.contains(|t| {
        matches!(
            t,
            Atomic::TFloat | Atomic::TIntegralFloat | Atomic::TLiteralFloat(..)
        )
    });
    if left_is_float || right_is_float {
        Type::single(Atomic::TFloat)
    } else if left.contains(coerces_to_int_in_arithmetic)
        && right.contains(coerces_to_int_in_arithmetic)
    {
        Type::single(Atomic::TInt)
    } else {
        let mut u = Type::empty();
        u.add_type(Atomic::TInt);
        u.add_type(Atomic::TFloat);
        u
    }
}

/// Type of the `/` operator. Unlike `+`/`-`/`*`, `int / int` yields `int|float` in PHP
/// because division may produce a fractional result (e.g. `5 / 2 = 2.5`).
pub fn infer_div(left: &Type, right: &Type) -> Type {
    if left.is_mixed() || right.is_mixed() {
        return Type::mixed();
    }
    let left_is_float = left.contains(|t| {
        matches!(
            t,
            Atomic::TFloat | Atomic::TIntegralFloat | Atomic::TLiteralFloat(..)
        )
    });
    let right_is_float = right.contains(|t| {
        matches!(
            t,
            Atomic::TFloat | Atomic::TIntegralFloat | Atomic::TLiteralFloat(..)
        )
    });
    if left_is_float || right_is_float {
        return Type::single(Atomic::TFloat);
    }
    let mut u = Type::empty();
    u.add_type(Atomic::TInt);
    u.add_type(Atomic::TFloat);
    u
}

/// Returns true when all atoms of `ty` produce a non-empty string in PHP's string cast.
///
/// Used by the concat (`.`) operator and `.=` assignment to determine whether the
/// result of a concatenation is guaranteed non-empty.
pub fn is_non_empty_when_concat(ty: &Type) -> bool {
    !ty.types.is_empty()
        && ty.types.iter().all(|a| match a {
            Atomic::TNonEmptyString
            | Atomic::TNumericString
            | Atomic::TCallableString
            | Atomic::TClassString(_)
            | Atomic::TInterfaceString(_)
            | Atomic::TEnumString
            | Atomic::TTraitString => true,
            Atomic::TLiteralString(s) => !s.is_empty(),
            // Any integer — including 0 — casts to a non-empty string ("0", "1", "-1", …)
            Atomic::TLiteralInt(_)
            | Atomic::TInt
            | Atomic::TPositiveInt
            | Atomic::TNegativeInt
            | Atomic::TNonNegativeInt
            | Atomic::TIntRange { .. } => true,
            // Any float casts to a non-empty string ("0", "1.5", …)
            Atomic::TFloat | Atomic::TIntegralFloat | Atomic::TLiteralFloat(..) => true,
            // true → "1"; false → "" so TBool and TFalse are excluded
            Atomic::TTrue => true,
            _ => false,
        })
}

/// Extract the string representation of a single scalar literal for concat folding.
/// Returns `None` for unions or non-literal types.
pub fn as_concat_str(ty: &Type) -> Option<String> {
    if ty.types.len() != 1 {
        return None;
    }
    match &ty.types[0] {
        Atomic::TLiteralString(s) => Some(s.as_ref().to_string()),
        Atomic::TLiteralInt(n) => Some(n.to_string()),
        Atomic::TTrue => Some("1".to_string()),
        Atomic::TFalse => Some(String::new()),
        _ => None,
    }
}

pub fn extract_simple_var(expr: &Expr) -> Option<String> {
    match &expr.kind {
        ExprKind::Variable(name) => Some(name.trim_start_matches('$').to_string()),
        ExprKind::Parenthesized(inner) => extract_simple_var(inner),
        _ => None,
    }
}

pub(crate) fn ast_params_to_fn_params_resolved(
    params: &[php_ast::owned::Param],
    self_fqcn: Option<&str>,
    db: &dyn crate::db::MirDatabase,
    file: &str,
) -> Vec<mir_codebase::DeclaredParam> {
    params
        .iter()
        .map(|p| {
            let name_str = p.name.as_deref().unwrap_or("").trim_start_matches('$');
            let ty = p
                .type_hint
                .as_ref()
                .map(|h| crate::parser::type_from_hint_owned(h, self_fqcn))
                .map(|u| resolve_named_objects_in_union(u, db, file));
            mir_codebase::DeclaredParam {
                name: Name::new(name_str),
                ty: mir_codebase::wrap_param_type(ty),
                out_ty: None,
                has_default: p.default.is_some(),
                is_variadic: p.variadic,
                is_byref: p.by_ref,
                is_optional: p.default.is_some() || p.variadic,
            }
        })
        .collect()
}

/// Merge `@param` docblock types into already-resolved closure/arrow-function params,
/// matching by parameter name.
///
/// Docblock types win over native hints — the same precedence used for top-level
/// function/method declarations — except when the native hint is a concrete scalar
/// whose family is entirely absent from the docblock type (e.g. `@param int $x` on a
/// `bool $x` hint), in which case the native hint is the runtime truth and wins.
pub(crate) fn apply_doc_param_types(
    params: &mut [mir_codebase::DeclaredParam],
    ast_params: &[php_ast::owned::Param],
    doc_params: &[(String, Type)],
    db: &dyn crate::db::MirDatabase,
    file: &str,
) {
    if doc_params.is_empty() {
        return;
    }
    for (param, ast_param) in params.iter_mut().zip(ast_params.iter()) {
        let name = ast_param
            .name
            .as_deref()
            .unwrap_or("")
            .trim_start_matches('$');
        let Some((_, doc_ty)) = doc_params.iter().find(|(n, _)| n == name) else {
            continue;
        };
        let mut doc_ty = resolve_named_objects_in_union(doc_ty.clone(), db, file);
        if let Some(native_ty) = param.ty.as_deref() {
            if crate::collector::native_hint_wins_over_docblock_scalar(native_ty, &doc_ty) {
                continue;
            }
            // Partial conflict: strip atoms foreign to the native hint's
            // scalar family instead of storing the raw docblock union.
            doc_ty = crate::collector::resolve_docblock_scalar_conflict(native_ty, doc_ty);
        }
        doc_ty.from_docblock = true;
        param.ty = mir_codebase::wrap_param_type(Some(doc_ty));
    }
}

pub(crate) fn resolve_named_objects_in_union(
    union: Type,
    db: &dyn crate::db::MirDatabase,
    file: &str,
) -> Type {
    let from_docblock = union.from_docblock;
    let possibly_undefined = union.possibly_undefined;
    let types: Vec<Atomic> = union
        .types
        .into_iter()
        .map(|a| resolve_named_objects_in_atomic(a, db, file))
        .collect();
    let mut result = Type::from_vec(types);
    result.from_docblock = from_docblock;
    result.possibly_undefined = possibly_undefined;
    result
}

/// Recurse into a type-argument list, array/list element+key type, or
/// intersection member — not just a top-level `TNamedObject` — so a
/// `use`-imported short name nested inside `Wrap<list<ShortName>>`,
/// `array<int, ShortName>`, or `ShortName&Other` is resolved the same as one
/// written at the top level.
fn resolve_named_objects_in_atomic(
    atomic: Atomic,
    db: &dyn crate::db::MirDatabase,
    file: &str,
) -> Atomic {
    match atomic {
        Atomic::TNamedObject { fqcn, type_params } => {
            let resolved = crate::db::resolve_name(db, file, fqcn.as_ref());
            let type_params = type_params
                .iter()
                .cloned()
                .map(|tp| resolve_named_objects_in_union(tp, db, file))
                .collect();
            Atomic::TNamedObject {
                fqcn: resolved.into(),
                type_params,
            }
        }
        Atomic::TArray { key, value } => Atomic::TArray {
            key: Box::new(resolve_named_objects_in_union(*key, db, file)),
            value: Box::new(resolve_named_objects_in_union(*value, db, file)),
        },
        Atomic::TList { value } => Atomic::TList {
            value: Box::new(resolve_named_objects_in_union(*value, db, file)),
        },
        Atomic::TIntersection { parts } => Atomic::TIntersection {
            parts: parts
                .iter()
                .cloned()
                .map(|p| resolve_named_objects_in_union(p, db, file))
                .collect(),
        },
        other => other,
    }
}

pub(crate) fn extract_string_from_expr(expr: &Expr) -> Option<String> {
    match &expr.kind {
        ExprKind::Identifier(s) => Some(s.trim_start_matches('$').to_string()),
        ExprKind::Variable(_) => None,
        ExprKind::String(s) => Some(s.to_string()),
        _ => None,
    }
}

/// For a literal `switch` case / `match` arm condition, return a
/// `(dedup_key, display)` pair. The key is type-tagged so that distinct
/// literal kinds never collide (e.g. the int `0` and the string `"0"`),
/// keeping duplicate detection free of PHP's loose-comparison surprises.
///
/// Returns `None` for any non-literal (variables, calls, negation, floats,
/// …) so dynamic conditions are never flagged — duplicate detection stays at
/// zero false positives.
fn literal_condition_key(expr: &Expr) -> Option<(String, String)> {
    match &expr.kind {
        ExprKind::Int(n) => Some((format!("int:{n}"), n.to_string())),
        ExprKind::String(s) => Some((format!("str:{s}"), format!("\"{s}\""))),
        ExprKind::Bool(b) => Some((format!("bool:{b}"), b.to_string())),
        ExprKind::Null => Some(("null".to_string(), "null".to_string())),
        _ => None,
    }
}

/// Given `switch`/`match` condition expressions in source order, return the
/// `(span, display)` of each literal whose value repeats an earlier one — the
/// duplicate branch can never be reached. Non-literal conditions are ignored,
/// so dynamic arms are never flagged.
pub fn duplicate_literal_conditions<'e>(
    conditions: impl Iterator<Item = &'e Expr>,
) -> Vec<(php_ast::Span, String)> {
    let mut seen = FxHashSet::default();
    let mut duplicates = Vec::new();
    for cond in conditions {
        if let Some((key, display)) = literal_condition_key(cond) {
            if !seen.insert(key) {
                duplicates.push((cond.span, display));
            }
        }
    }
    duplicates
}

/// Returns true if `ty` contains any reference to a template param name from `names`,
/// including names nested inside generic type arguments (e.g. `R` inside `Result<Throwable, R>`).
/// Handles both `TTemplateParam` and the docblock-parser workaround where bare unqualified names
/// are emitted as `TNamedObject { fqcn: "T", type_params: [] }`.
pub(crate) fn type_refs_any_template(ty: &Type, names: &FxHashSet<Name>) -> bool {
    fn check_atomic(a: &Atomic, names: &FxHashSet<Name>) -> bool {
        match a {
            Atomic::TTemplateParam { name, .. } => names.contains(name),
            Atomic::TNamedObject { fqcn, type_params } => {
                if type_params.is_empty() && !fqcn.contains('\\') && names.contains(fqcn) {
                    return true;
                }
                type_params
                    .iter()
                    .any(|tp| tp.types.iter().any(|a| check_atomic(a, names)))
            }
            Atomic::TClassString(Some(inner)) => !inner.contains('\\') && names.contains(inner),
            _ => false,
        }
    }
    ty.types.iter().any(|a| check_atomic(a, names))
}

fn scalar_types_compatible(value_ty: &Type, prop_ty: &Type) -> bool {
    value_ty.is_subtype_structural(prop_ty)
}

pub(crate) fn property_assign_compatible(
    value_ty: &Type,
    prop_ty: &Type,
    db: &dyn crate::db::MirDatabase,
) -> bool {
    if scalar_types_compatible(value_ty, prop_ty) {
        return true;
    }
    if is_subtype(db, value_ty, prop_ty) {
        return true;
    }
    value_ty.types.iter().all(|a| match a {
        Atomic::TTemplateParam { .. } => true,
        Atomic::TClosure { .. } | Atomic::TCallable { .. } => prop_ty.types.iter().any(|p| {
            matches!(p, Atomic::TClosure { .. } | Atomic::TCallable { .. })
                || matches!(p, Atomic::TNamedObject { fqcn, .. } if fqcn.as_ref() == "Closure")
        }),
        Atomic::TNever => true,
        Atomic::TNull => prop_ty.is_nullable(),
        _ => false,
    })
}

pub(crate) fn is_property_type_coercion(
    value_ty: &Type,
    prop_ty: &Type,
    db: &dyn crate::db::MirDatabase,
) -> bool {
    if value_ty.is_mixed() || prop_ty.is_mixed() {
        return false;
    }
    let value_core = value_ty.core_type();
    if value_core.types.is_empty() || !value_core.is_single() {
        return false;
    }
    let val_fqcn = match value_core.types.first().unwrap() {
        Atomic::TNamedObject { fqcn, type_params } if type_params.is_empty() => *fqcn,
        _ => return false,
    };
    prop_ty.types.iter().any(|p| {
        let prop_fqcn = match p {
            Atomic::TNamedObject { fqcn, type_params } if type_params.is_empty() => fqcn,
            _ => return false,
        };
        crate::db::extends_or_implements(db, prop_fqcn.as_ref(), val_fqcn.as_ref())
    })
}

#[cfg(test)]
mod range_arithmetic_tests {
    use super::*;

    fn range(min: Option<i64>, max: Option<i64>) -> Type {
        Type::single(Atomic::TIntRange { min, max })
    }

    fn lit(n: i64) -> Type {
        Type::single(Atomic::TLiteralInt(n))
    }

    #[test]
    fn add_shifts_both_bounds() {
        // int<0, 4> + 5  =>  int<5, 9>
        let r =
            infer_int_range_arithmetic(&range(Some(0), Some(4)), &lit(5), BinaryOp::Add).unwrap();
        assert_eq!(r.to_string(), "int<5, 9>");
    }

    #[test]
    fn add_keeps_unbounded_upper() {
        // int<0, max> + 5  =>  int<5, max>
        let r = infer_int_range_arithmetic(&range(Some(0), None), &lit(5), BinaryOp::Add).unwrap();
        assert_eq!(r.to_string(), "int<5, max>");
    }

    #[test]
    fn sub_lowers_min_to_negative() {
        // int<0, max> - 1  =>  int<-1, max>   (lmin - rmax, lmax - rmin)
        let r = infer_int_range_arithmetic(&range(Some(0), None), &lit(1), BinaryOp::Sub).unwrap();
        assert_eq!(r.to_string(), "int<-1, max>");
    }

    #[test]
    fn add_overflow_saturates_to_unbounded() {
        // int<i64::MAX, i64::MAX> + 1  =>  both bounds overflow to unbounded,
        // which renders as the bare `int`.
        let r = infer_int_range_arithmetic(
            &range(Some(i64::MAX), Some(i64::MAX)),
            &lit(1),
            BinaryOp::Add,
        )
        .unwrap();
        assert_eq!(r.to_string(), "int");
    }

    #[test]
    fn no_range_operand_returns_none() {
        // plain int + literal: no explicit range, so range arithmetic abstains
        assert!(
            infer_int_range_arithmetic(&Type::single(Atomic::TInt), &lit(3), BinaryOp::Add)
                .is_none()
        );
    }

    #[test]
    fn non_integer_operand_returns_none() {
        // range + string: not integer-only, abstain
        assert!(infer_int_range_arithmetic(
            &range(Some(0), None),
            &Type::single(Atomic::TString),
            BinaryOp::Add
        )
        .is_none());
    }

    #[test]
    fn mul_non_negative_ranges() {
        // non-negative × literal positive → int<0, max> (unbounded above)
        let r = infer_int_range_arithmetic(&range(Some(0), None), &lit(2), BinaryOp::Mul).unwrap();
        assert_eq!(r, range(Some(0), None));

        // bounded × bounded → bounded product
        let r = infer_int_range_arithmetic(
            &range(Some(2), Some(4)),
            &range(Some(3), Some(6)),
            BinaryOp::Mul,
        )
        .unwrap();
        assert_eq!(r, range(Some(6), Some(24)));

        // mixed-sign operand: defer to infer_arithmetic
        assert!(
            infer_int_range_arithmetic(&range(None, Some(-1)), &lit(2), BinaryOp::Mul).is_none()
        );
    }

    #[test]
    fn mod_non_negative_ranges() {
        // non-negative-int % 5 → int<0, 4>
        let r = infer_int_range_arithmetic(&range(Some(0), None), &lit(5), BinaryOp::Mod).unwrap();
        assert_eq!(r, range(Some(0), Some(4)));

        // int<0, 100> % 10 → int<0, 9>
        let r = infer_int_range_arithmetic(&range(Some(0), Some(100)), &lit(10), BinaryOp::Mod)
            .unwrap();
        assert_eq!(r, range(Some(0), Some(9)));

        // negative divisor: no range inference
        assert!(
            infer_int_range_arithmetic(&range(Some(0), None), &lit(-5), BinaryOp::Mod).is_none()
        );
    }
}