jsonschema 0.49.2

JSON schema validaton library
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
#[cfg(feature = "arbitrary-precision")]
use std::borrow::Cow;
use std::{cell::RefCell, fmt, io, mem};

#[cfg(feature = "arbitrary-precision")]
use serde::ser::SerializeStruct;
use serde::{
    ser::{self, Serialize, SerializeMap, SerializeSeq},
    Serializer,
};
use serde_json::{
    ser::{CompactFormatter, Formatter},
    Number, Value,
};

const I64_UPPER_EXCLUSIVE_F64: f64 = 9_223_372_036_854_775_808.0;
const I64_LOWER_INCLUSIVE_F64: f64 = -9_223_372_036_854_775_808.0;
const U64_UPPER_EXCLUSIVE_F64: f64 = 18_446_744_073_709_551_616.0;
const RECURSION_LIMIT: u16 = 255;
const MAX_SCRATCH_POOL_SIZE: usize = 8;
const MAX_SCRATCH_CAPACITY: usize = 16_384;
#[cfg(feature = "arbitrary-precision")]
const SERDE_JSON_NUMBER_TOKEN: &str = "$serde_json::private::Number";
#[cfg(feature = "arbitrary-precision")]
const MAX_EXPANDED_INTEGER_DIGITS: usize = 1 << 20;

/// An integer-valued `f64` within the `u64` range as `u64`, else `None`.
#[expect(
    clippy::cast_possible_truncation,
    clippy::cast_sign_loss,
    reason = "guarded by the `0.0..U64_UPPER_EXCLUSIVE_F64` range and zero fractional part"
)]
pub(crate) fn integer_valued_u64(value: f64) -> Option<u64> {
    (value.fract() == 0.0 && (0.0..U64_UPPER_EXCLUSIVE_F64).contains(&value))
        .then_some(value as u64)
}

/// An integer-valued `f64` within the `i64` range as `i64`, else `None`.
#[expect(
    clippy::cast_possible_truncation,
    reason = "guarded by the `I64_LOWER_INCLUSIVE_F64..I64_UPPER_EXCLUSIVE_F64` range and zero fractional part"
)]
pub(crate) fn integer_valued_i64(value: f64) -> Option<i64> {
    (value.fract() == 0.0 && (I64_LOWER_INCLUSIVE_F64..I64_UPPER_EXCLUSIVE_F64).contains(&value))
        .then_some(value as i64)
}

/// Error returned by [`to_string`].
#[derive(Debug)]
pub struct Error {
    inner: serde_json::Error,
}

impl fmt::Display for Error {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        self.inner.fmt(f)
    }
}

impl std::error::Error for Error {
    fn source(&self) -> Option<&(dyn std::error::Error + 'static)> {
        Some(&self.inner)
    }
}

impl From<serde_json::Error> for Error {
    fn from(inner: serde_json::Error) -> Self {
        Error { inner }
    }
}

/// Serialize a JSON value into a deterministic canonical JSON string, so equivalent values share one text.
///
/// # Rules
///
/// - Object keys are emitted in lexicographic order.
/// - Integer-valued floats are emitted as integers (`1.0` becomes `1`).
/// - With `arbitrary-precision`, one normal form per value: plain integer/decimal while the expansion fits the digit
///   cap, normalized scientific (`d[.rest]e{E}`) past it. The IR relies on this for value text equality and type scanning.
/// - Output is always compact (no extra whitespace).
///
/// # Examples
///
/// ```rust
/// use serde_json::json;
///
/// let schema = json!({"b": 1, "a": {"y": 2, "x": 3}});
/// let canonical = jsonschema::canonical::json::to_string(&schema).unwrap();
/// assert_eq!(canonical, r#"{"a":{"x":3,"y":2},"b":1}"#);
/// ```
///
/// # Errors
///
/// Returns an error if serialization fails, e.g. when the input exceeds the recursion limit.
#[allow(
    clippy::missing_panics_doc,
    reason = "the canonical serializer emits valid UTF-8 by construction; the expect is unreachable"
)]
pub fn to_string(value: &Value) -> Result<String, Error> {
    let mut output = Vec::with_capacity(initial_output_capacity(value));
    let formatter = CanonicalFormatter {
        default: CompactFormatter,
    };
    let scratch_pool = RefCell::new(Vec::new());
    let mut serializer = serde_json::Serializer::with_formatter(&mut output, formatter);
    CanonicalValue::new(value, 0, &scratch_pool)
        .serialize(&mut serializer)
        .map_err(Error::from)?;
    Ok(String::from_utf8(output).expect("canonical serializer emits valid UTF-8"))
}

#[inline]
fn initial_output_capacity(value: &Value) -> usize {
    const MIN_CAPACITY: usize = 16;
    const MAX_PREALLOC: usize = 1 << 20; // 1 MiB

    let estimated = match value {
        Value::Object(map) => map.len().saturating_mul(24).saturating_add(2),
        Value::Array(items) => items.len().saturating_mul(12).saturating_add(2),
        Value::String(s) => s.len().saturating_add(2),
        Value::Number(_) => 32,
        Value::Bool(_) => 8,
        Value::Null => 4,
    };

    estimated.clamp(MIN_CAPACITY, MAX_PREALLOC)
}

#[derive(Default)]
struct CanonicalFormatter {
    default: CompactFormatter,
}

/// A formatter that emits integer-valued floats as integers.
impl Formatter for CanonicalFormatter {
    #[inline]
    fn write_f64<W: io::Write + ?Sized>(&mut self, writer: &mut W, value: f64) -> io::Result<()> {
        if let Some(integer) = integer_valued_u64(value) {
            return self.default.write_u64(writer, integer);
        }
        if let Some(integer) = integer_valued_i64(value) {
            return self.default.write_i64(writer, integer);
        }
        if value.fract() == 0.0 {
            return writer.write_all(format!("{value:.0}").as_bytes());
        }
        self.default.write_f64(writer, value)
    }
}

struct CanonicalValue<'value> {
    value: &'value Value,
    recursion_depth: u16,
    scratch_pool: &'value RefCell<Vec<Vec<ObjectEntry<'value>>>>,
}

struct ObjectEntry<'value> {
    key: &'value str,
    value: &'value Value,
}

struct ObjectEntryScratch<'value, 'pool> {
    entries: Vec<ObjectEntry<'value>>,
    pool: &'pool RefCell<Vec<Vec<ObjectEntry<'value>>>>,
}

impl<'value, 'pool> ObjectEntryScratch<'value, 'pool> {
    fn with_capacity(pool: &'pool RefCell<Vec<Vec<ObjectEntry<'value>>>>, capacity: usize) -> Self {
        let mut entries = pool.borrow_mut().pop().unwrap_or_default();
        entries.clear();
        if entries.capacity() < capacity {
            entries.reserve(capacity - entries.capacity());
        }
        Self { entries, pool }
    }

    #[inline]
    fn entries_mut(&mut self) -> &mut Vec<ObjectEntry<'value>> {
        &mut self.entries
    }

    #[inline]
    fn entries(&self) -> &[ObjectEntry<'value>] {
        &self.entries
    }
}

impl Drop for ObjectEntryScratch<'_, '_> {
    fn drop(&mut self) {
        self.entries.clear();
        if self.entries.capacity() > MAX_SCRATCH_CAPACITY {
            return;
        }
        let mut pool = self.pool.borrow_mut();
        if pool.len() < MAX_SCRATCH_POOL_SIZE {
            pool.push(mem::take(&mut self.entries));
        }
    }
}

impl<'value> CanonicalValue<'value> {
    #[inline]
    const fn new(
        value: &'value Value,
        recursion_depth: u16,
        scratch_pool: &'value RefCell<Vec<Vec<ObjectEntry<'value>>>>,
    ) -> Self {
        CanonicalValue {
            value,
            recursion_depth,
            scratch_pool,
        }
    }
}

/// Emits already-canonical number text verbatim, without parsing it into a [`Number`].
#[cfg(feature = "arbitrary-precision")]
#[doc(hidden)]
pub struct BorrowedNumber<'a>(pub &'a str);

#[cfg(feature = "arbitrary-precision")]
impl Serialize for BorrowedNumber<'_> {
    #[inline]
    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
    where
        S: Serializer,
    {
        let mut serialized = serializer.serialize_struct(SERDE_JSON_NUMBER_TOKEN, 1)?;
        serialized.serialize_field(SERDE_JSON_NUMBER_TOKEN, self.0)?;
        serialized.end()
    }
}

#[cfg(feature = "arbitrary-precision")]
fn push_digits(output: &mut String, digits: &[u8]) {
    output.push_str(std::str::from_utf8(digits).expect("ASCII digits"));
}

/// Canonical text for a valid JSON-number token, borrowing the input when it is already canonical.
/// `None` only when the exponent magnitude overflows `usize` (32-bit targets), leaving the raw text.
#[cfg(feature = "arbitrary-precision")]
#[doc(hidden)]
#[inline]
#[must_use]
pub fn canonical_number(raw: &str) -> Option<Cow<'_, str>> {
    // `raw` is a valid JSON-number token; the scan assumes that shape.
    let bytes = raw.as_bytes();

    let mut idx = 0;
    let negative = bytes[idx] == b'-';
    if negative {
        idx += 1;
    }

    let integer_start = idx;
    if bytes[idx] == b'0' {
        idx += 1;
    } else {
        while idx < bytes.len() && bytes[idx].is_ascii_digit() {
            idx += 1;
        }
    }
    let integer_end = idx;

    let mut fraction_start = idx;
    let mut fraction_end = idx;
    if idx < bytes.len() && bytes[idx] == b'.' {
        idx += 1;
        fraction_start = idx;
        while idx < bytes.len() && bytes[idx].is_ascii_digit() {
            idx += 1;
        }
        fraction_end = idx;
    }
    let has_fraction = fraction_end > fraction_start;

    let mut exponent: i64 = 0;
    let mut exponent_digits = "";
    let has_exponent = if idx < bytes.len() && matches!(bytes[idx], b'e' | b'E') {
        idx += 1;
        // Keep a leading `-` for `parse_bytes`; drop `+`.
        let sign_start = idx;
        let mut exponent_negative = false;
        if idx < bytes.len() && matches!(bytes[idx], b'+' | b'-') {
            exponent_negative = bytes[idx] == b'-';
            idx += 1;
        }
        let digits_start = idx;
        while idx < bytes.len() && bytes[idx].is_ascii_digit() {
            // Saturates: steers the branch below; scientific form recomputes exactly.
            exponent = exponent
                .saturating_mul(10)
                .saturating_add(i64::from(bytes[idx] - b'0'));
            idx += 1;
        }
        exponent_digits = if exponent_negative {
            exponent = -exponent;
            &raw[sign_start..idx]
        } else {
            &raw[digits_start..idx]
        };
        true
    } else {
        false
    };

    let integer_within_cap = integer_end - integer_start <= MAX_EXPANDED_INTEGER_DIGITS;

    if !has_fraction && !has_exponent && integer_within_cap {
        return Some(Cow::Borrowed(raw));
    }

    // In-cap plain decimal with a non-zero last digit is already canonical.
    if !has_exponent && integer_within_cap && bytes[fraction_end - 1] != b'0' {
        return Some(Cow::Borrowed(raw));
    }

    let integer_digits = &raw[integer_start..integer_end];
    let fraction_digits = &raw[fraction_start..fraction_end];
    let mut digits = Vec::with_capacity(integer_digits.len() + fraction_digits.len());
    digits.extend_from_slice(integer_digits.as_bytes());
    digits.extend_from_slice(fraction_digits.as_bytes());

    // Strip leading zeros so the cap check below can't split equal values across branches.
    let leading_zeros = digits.iter().take_while(|&&byte| byte == b'0').count();
    digits.drain(..leading_zeros);
    if digits.is_empty() {
        return Some(Cow::Borrowed("0"));
    }

    let parts = NumberParts {
        fraction_len: fraction_digits.len(),
        exponent_digits,
        negative,
    };

    let fraction_len = i64::try_from(parts.fraction_len).unwrap_or(i64::MAX);
    let shift = exponent.saturating_sub(fraction_len);

    let prefix_len = if shift >= 0 {
        let extra_zeros = usize::try_from(shift).ok()?;
        let expanded_len = digits.len().checked_add(extra_zeros)?;
        if expanded_len > MAX_EXPANDED_INTEGER_DIGITS {
            return canonical_scientific_number(&digits, &parts).map(Cow::Owned);
        }
        digits.resize(expanded_len, b'0');
        digits.len()
    } else {
        let drop_len = usize::try_from(shift.unsigned_abs()).ok()?;
        if drop_len <= digits.len()
            && digits[digits.len() - drop_len..]
                .iter()
                .all(|&byte| byte == b'0')
        {
            digits.len() - drop_len
        } else {
            return canonical_fractional_number(&digits, drop_len, &parts).map(Cow::Owned);
        }
    };

    // Prefix is non-empty with no leading zero.
    let prefix = &digits[..prefix_len];
    let mut output = String::with_capacity(prefix.len() + usize::from(negative));
    if negative {
        output.push('-');
    }
    push_digits(&mut output, prefix);
    Some(Cow::Owned(output))
}

/// Parsed number facets needed past the digit buffer: fraction length, the exact exponent digits (the
/// `i64` working exponent saturates; the slice keeps a leading `-`), and the sign.
#[cfg(feature = "arbitrary-precision")]
struct NumberParts<'a> {
    fraction_len: usize,
    exponent_digits: &'a str,
    negative: bool,
}

/// Plain-decimal normal form: `digits` (leading zeros pre-stripped) sit `point_offset` places right of the
/// point. Trailing zeros are stripped before the `MAX_EXPANDED_INTEGER_DIGITS` cap check so equal values never
/// straddle it; past the cap the scientific form takes over.
#[cfg(feature = "arbitrary-precision")]
fn canonical_fractional_number(
    digits: &[u8],
    point_offset: usize,
    parts: &NumberParts<'_>,
) -> Option<String> {
    let last_non_zero = digits.iter().rposition(|&byte| byte != b'0')?;
    // Stripped trailing zeros sit right of the point, so the offset stays positive.
    let point_offset = point_offset.checked_sub(digits.len() - 1 - last_non_zero)?;
    if point_offset > MAX_EXPANDED_INTEGER_DIGITS {
        return canonical_scientific_number(digits, parts);
    }
    let stripped = &digits[..=last_non_zero];
    let mut output = String::with_capacity(
        stripped
            .len()
            .max(point_offset)
            .saturating_add(2)
            .saturating_add(usize::from(parts.negative)),
    );
    if parts.negative {
        output.push('-');
    }
    if stripped.len() > point_offset {
        let integer_len = stripped.len() - point_offset;
        push_digits(&mut output, &stripped[..integer_len]);
        output.push('.');
        push_digits(&mut output, &stripped[integer_len..]);
    } else {
        output.push_str("0.");
        output.extend(std::iter::repeat_n('0', point_offset - stripped.len()));
        push_digits(&mut output, stripped);
    }
    Some(output)
}

/// Scientific normal form `d[.rest]e{E}` for values whose plain expansion exceeds `MAX_EXPANDED_INTEGER_DIGITS`.
/// `digits` carry `digits x 10^(exponent - fraction_len)`, so `E = exponent + len(digits) - fraction_len - 1`.
#[cfg(feature = "arbitrary-precision")]
fn canonical_scientific_number(digits: &[u8], parts: &NumberParts<'_>) -> Option<String> {
    use num_bigint::BigInt;

    // The plain-decimal re-dispatch has no exponent; `parse_bytes` on an empty slice gives `None`.
    let mut exponent = if parts.exponent_digits.is_empty() {
        BigInt::from(0)
    } else {
        BigInt::parse_bytes(parts.exponent_digits.as_bytes(), 10)?
    };
    exponent += BigInt::from(i64::try_from(digits.len()).ok()?);
    exponent -= BigInt::from(i64::try_from(parts.fraction_len).ok()?) + BigInt::from(1);

    let last_non_zero = digits.iter().rposition(|&byte| byte != b'0')?;
    let significand = &digits[..=last_non_zero];
    let exponent = exponent.to_string();
    let mut output = String::with_capacity(
        significand
            .len()
            .saturating_add(exponent.len())
            .saturating_add(3),
    );
    if parts.negative {
        output.push('-');
    }
    output.push(char::from(significand[0]));
    if significand.len() > 1 {
        output.push('.');
        push_digits(&mut output, &significand[1..]);
    }
    output.push('e');
    output.push_str(&exponent);
    Some(output)
}

// `Number::from_f64` formats to shortest-roundtrip text before this runs, so the double nearest
// `1e300` and exactly `1e300` collide here. The bindings convert native floats exactly and do not.
fn serialize_number<S>(number: &Number, serializer: S) -> Result<S::Ok, S::Error>
where
    S: Serializer,
{
    #[cfg(feature = "arbitrary-precision")]
    {
        if let Some(text) = canonical_number(number.as_str()) {
            return serializer.serialize_some(&BorrowedNumber(&text));
        }
    }
    number.serialize(serializer)
}

impl Serialize for CanonicalValue<'_> {
    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
    where
        S: Serializer,
    {
        match self.value {
            Value::Null => serializer.serialize_unit(),
            Value::Bool(value) => serializer.serialize_bool(*value),
            Value::Number(number) => serialize_number(number, serializer),
            Value::String(value) => serializer.serialize_str(value),
            Value::Array(items) => {
                if self.recursion_depth == RECURSION_LIMIT {
                    return Err(ser::Error::custom("Recursion limit reached"));
                }
                let mut sequence = serializer.serialize_seq(Some(items.len()))?;
                for item in items {
                    sequence.serialize_element(&CanonicalValue::new(
                        item,
                        self.recursion_depth + 1,
                        self.scratch_pool,
                    ))?;
                }
                sequence.end()
            }
            Value::Object(map) => {
                if self.recursion_depth == RECURSION_LIMIT {
                    return Err(ser::Error::custom("Recursion limit reached"));
                }
                let mut output = serializer.serialize_map(Some(map.len()))?;
                // Always sort keys: downstream crates can enable `serde_json/preserve_order` transitively.
                let mut scratch = ObjectEntryScratch::with_capacity(self.scratch_pool, map.len());
                {
                    let entries = scratch.entries_mut();
                    for (key, value) in map {
                        entries.push(ObjectEntry {
                            key: key.as_str(),
                            value,
                        });
                    }
                    entries.sort_unstable_by(|left, right| {
                        left.key.as_bytes().cmp(right.key.as_bytes())
                    });
                }
                for entry in scratch.entries() {
                    output.serialize_entry(
                        entry.key,
                        &CanonicalValue::new(
                            entry.value,
                            self.recursion_depth + 1,
                            self.scratch_pool,
                        ),
                    )?;
                }
                output.end()
            }
        }
    }
}

#[cfg(test)]
mod tests {
    use serde_json::{json, Map, Value};
    use test_case::test_case;

    use super::to_string;

    #[test]
    fn canonical_string_is_stable_for_equivalent_schemas() {
        let left: Value =
            serde_json::from_str(r#"{"b":1,"a":{"z":3,"x":1,"y":2},"c":[{"d":4,"b":2}]}"#).unwrap();
        let right: Value =
            serde_json::from_str(r#"{"c":[{"b":2,"d":4}],"a":{"y":2,"x":1,"z":3},"b":1}"#).unwrap();

        assert_eq!(to_string(&left).unwrap(), to_string(&right).unwrap());
    }

    #[test_case("null"; "null")]
    #[test_case("true"; "bool_true")]
    #[test_case("false"; "bool_false")]
    #[test_case(r#""hello""#; "simple")]
    #[test_case(r#""line\nbreak""#; "escaped_newline")]
    fn scalar_literals_roundtrip(raw: &str) {
        let value: Value = serde_json::from_str(raw).unwrap();
        assert_eq!(to_string(&value).unwrap(), raw);
    }

    #[test]
    fn canonical_output_is_idempotent() {
        let value: Value =
            serde_json::from_str(r#"{"z":{"b":1,"a":2},"a":[3,2,1],"f":1.0,"v":1.5}"#).unwrap();

        let first = to_string(&value).unwrap();
        let parsed: Value = serde_json::from_str(&first).unwrap();

        assert_eq!(to_string(&parsed).unwrap(), first);
    }

    fn nest_arrays(depth: usize) -> Value {
        let mut value = Value::Null;
        for _ in 0..depth {
            value = Value::Array(vec![value]);
        }
        value
    }

    fn nest_objects(depth: usize) -> Value {
        let mut value = Value::Null;
        for _ in 0..depth {
            value = Value::Object([("a".to_string(), value)].into_iter().collect());
        }
        value
    }

    #[test_case(&nest_arrays(300) ; "arrays")]
    #[test_case(&nest_objects(300) ; "objects")]
    fn recursion_limit_error_exposes_source_and_message(value: &Value) {
        let error = to_string(value).expect_err("should exceed the recursion limit");
        assert!(std::error::Error::source(&error).is_some());
        assert!(error.to_string().contains("Recursion"));
    }

    #[test]
    fn oversized_object_scratch_is_not_pooled() {
        // A scratch buffer whose capacity exceeds MAX_SCRATCH_CAPACITY is dropped instead of pooled.
        let mut map = Map::new();
        for index in 0..20_000 {
            map.insert(format!("k{index:05}"), json!(index));
        }
        let out = to_string(&Value::Object(map)).unwrap();
        assert!(out.starts_with(r#"{"k00000":0,"k00001":1,"#));
    }

    #[cfg(not(feature = "arbitrary-precision"))]
    #[test_case(&json!(1.0), "1" ; "small unsigned integral")]
    #[test_case(&json!(-5.0), "-5" ; "small signed integral")]
    #[test_case(&json!(1.5), "1.5" ; "non integral falls through")]
    fn write_f64_formats(value: &Value, expected: &str) {
        assert_eq!(to_string(value).unwrap(), expected);
    }

    #[cfg(not(feature = "arbitrary-precision"))]
    #[test_case(1e30 ; "above u64 max")]
    #[test_case(-1e30 ; "below i64 min")]
    fn large_integral_floats_have_no_exponent(value: f64) {
        let out = to_string(&json!(value)).unwrap();
        assert!(
            !out.contains('.') && !out.contains('e') && !out.contains('E'),
            "{out}"
        );
    }

    #[test]
    fn large_integer_valued_float_uses_integer_form() {
        let value: Value = serde_json::from_str("1e300").unwrap();

        #[cfg(feature = "arbitrary-precision")]
        let expected = {
            let mut output = String::with_capacity(301);
            output.push('1');
            output.push_str(&"0".repeat(300));
            output
        };
        #[cfg(not(feature = "arbitrary-precision"))]
        let expected = format!("{:.0}", 1e300_f64);

        assert_eq!(to_string(&value).unwrap(), expected);
    }

    #[cfg(feature = "arbitrary-precision")]
    fn canonical(raw: &str) -> String {
        to_string(&serde_json::from_str::<Value>(raw).unwrap()).unwrap()
    }

    // Plain expansion past the digit cap switches to scientific normal form.
    #[cfg(feature = "arbitrary-precision")]
    #[test_case("1e1048577", "1e1048577" ; "single digit significand")]
    #[test_case("12e1048576", "1.2e1048577" ; "multi digit significand")]
    #[test_case("-1e1048576", "-1e1048576" ; "negative significand")]
    // Huge negative exponent drops past the digit buffer into the fractional-then-scientific path.
    #[test_case("1e-1048577", "1e-1048577" ; "fractional past cap")]
    // All-zero fractions collapse to "0"; trailing-zero fractions expand back to a (signed) integer.
    #[test_case("0.0", "0" ; "all zero collapses")]
    #[test_case("-1.0", "-1" ; "negative integral expansion")]
    fn scientific_normal_form(raw: &str, expected: &str) {
        assert_eq!(canonical(raw), expected);
    }

    // Within-cap spellings share one plain-decimal normal form.
    #[cfg(feature = "arbitrary-precision")]
    #[test_case("1e+3", "1000" ; "positive exponent expands")]
    #[test_case("12e2", "1200" ; "multi digit positive exponent expands")]
    #[test_case("100E-2", "1" ; "negative exponent integral")]
    #[test_case("1e-2", "0.01" ; "negative exponent zero padded")]
    #[test_case("3.14e-3", "0.00314" ; "fraction with exponent zero padded")]
    #[test_case("3.1400e-3", "0.00314" ; "trailing zeros stripped before padding")]
    #[test_case("1.50", "1.5" ; "trailing fraction zero stripped")]
    #[test_case("0.15e1", "1.5" ; "point shifts right")]
    #[test_case("-5e-1", "-0.5" ; "negative fraction")]
    #[test_case("0e+10", "0" ; "zero with positive exponent")]
    #[test_case("-0E-1000", "0" ; "negative zero with huge negative exponent")]
    // No exponent, already canonical, non-zero trailing digit: takes the zero-copy borrow path.
    #[test_case("1.25", "1.25" ; "fractional decimal")]
    fn plain_decimal_normal_form(raw: &str, expected: &str) {
        assert_eq!(canonical(raw), expected);
    }

    // A plain decimal past the digit cap re-dispatches to scientific form with no explicit exponent,
    // sharing one text with the equivalent scientific spelling. The trailing `0` defeats the
    // already-canonical fast path so the value reaches the expansion cap.
    #[cfg(feature = "arbitrary-precision")]
    #[test]
    fn oversized_plain_decimal_matches_scientific() {
        let plain = format!("0.{}10", "0".repeat(1 << 20)); // 10^-1048577
        assert_eq!(canonical(&plain), "1e-1048577");
        assert_eq!(canonical(&plain), canonical("1e-1048577"));
    }
}