rustledger-core 0.22.0

Core types for rustledger: Amount, Position, Inventory, and all directive types
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
//! Decimal arithmetic with Python `decimal` scale semantics.
//!
//! `rust_decimal` and Python's `decimal` agree on the VALUE of a sum but not
//! always on its SCALE, and BQL renders a naked decimal at its intrinsic scale
//! (bean-query's `DecimalRenderer` pads for alignment only — it never
//! quantizes). So a scale difference is a visible output difference.
//!
//! Python's rule for `+`/`-` is that an exact result carries
//! `max(scale(a), scale(b))`:
//!
//! ```text
//!   Decimal("0.00") + Decimal("1")  ==  Decimal("1.00")
//! ```
//!
//! `rust_decimal` matches that for ordinary operands (`2.00 + 1 == 3.00`) but
//! not when one side is ZERO: its addition returns the other operand
//! unchanged, so the zero's scale is discarded and `0.00 + 1 == 1`.
//!
//! That makes an accumulation order-dependent, which is how it surfaced. A
//! `SUM` over `-1, 1, -111.11, 111.11, -2, 2` passes through `0.00` at the
//! fourth term; the `-2` that follows then resets the running total to scale
//! 0 and the query prints `0` where bean-query prints `0.00`. Move the
//! fractional pair last and the same multiset prints `0.00` — same value,
//! same inputs, different rendering.

use rust_decimal::Decimal;

/// Add two decimals with Python `decimal`'s scale rule.
///
/// The value is `a + b` either way; this only restores the scale
/// `rust_decimal` drops when an operand is zero (see the module docs). The
/// result is padded UP to `max(scale(a), scale(b))` and never truncated, so
/// it cannot lose significant digits.
///
/// Rescaling is best-effort: `Decimal::rescale` is a no-op when the target
/// scale would overflow the 96-bit mantissa (a value near `Decimal::MAX` at
/// high scale). Such a value cannot be represented at that scale at all, so
/// there is nothing to restore and the unrescaled sum is the best available
/// answer — the same one we returned before this function existed.
#[must_use]
pub fn add_python_scale(a: Decimal, b: Decimal) -> Decimal {
    let mut sum = a + b;
    let target = a.scale().max(b.scale());
    if sum.scale() < target {
        sum.rescale(target);
    }
    sum
}

/// Subtract with Python `decimal`'s scale rule — see [`add_python_scale`].
#[must_use]
pub fn sub_python_scale(a: Decimal, b: Decimal) -> Decimal {
    let mut diff = a - b;
    let target = a.scale().max(b.scale());
    if diff.scale() < target {
        diff.rescale(target);
    }
    diff
}

/// Overflow-checked [`add_python_scale`].
///
/// `None` on value-range overflow, matching `Decimal::checked_add` — BQL maps
/// that to NULL rather than panicking, so the checked form is what the
/// expression evaluator needs.
#[must_use]
pub fn checked_add_python_scale(a: Decimal, b: Decimal) -> Option<Decimal> {
    let mut sum = a.checked_add(b)?;
    let target = a.scale().max(b.scale());
    if sum.scale() < target {
        sum.rescale(target);
    }
    Some(sum)
}

/// Overflow-checked [`sub_python_scale`] — see [`checked_add_python_scale`].
#[must_use]
pub fn checked_sub_python_scale(a: Decimal, b: Decimal) -> Option<Decimal> {
    let mut diff = a.checked_sub(b)?;
    let target = a.scale().max(b.scale());
    if diff.scale() < target {
        diff.rescale(target);
    }
    Some(diff)
}

/// Divide with Python `decimal`'s scale rule.
///
/// Python defines an *ideal exponent* for division: `exp(a) - exp(b)`, i.e.
/// an ideal SCALE of `scale(a) - scale(b)`. An exact quotient is reduced
/// toward that scale but never below the scale exactness requires:
///
/// ```text
///   0.00 / 4     ->  0.00     ideal 2, exact needs 0  -> 2
///   7    / 2     ->  3.5      ideal 0, exact needs 1  -> 1
///   1.00 / 2     ->  0.50     ideal 2, exact needs 1  -> 2
///   1.000 / 8    ->  0.125    ideal 3, exact needs 3  -> 3
/// ```
///
/// `rust_decimal` misses this in BOTH directions, so a pad-only fix would be
/// half a fix:
///
/// * **Under**, on a zero dividend — the same zero shortcut behind
///   [`add_python_scale`]. `0.00 / 4` gives `0`, dropping the scale.
/// * **Over**, on an exact quotient — `7 / 2` gives `3.50`, one trailing
///   zero more than the ideal exponent allows.
///
/// So the quotient is stripped to its minimal form and then padded up to the
/// ideal scale; the two steps together land on Python's answer from either
/// side. An inexact quotient (`1 / 3`) has no trailing zeros to strip and a
/// scale far past the ideal, so both steps are no-ops and it is returned as
/// `rust_decimal` computed it.
///
/// Returns `None` on divide-by-zero or overflow, matching
/// `Decimal::checked_div` — BQL maps that to NULL rather than panicking.
#[must_use]
pub fn checked_div_python_scale(a: Decimal, b: Decimal) -> Option<Decimal> {
    let quotient = a.checked_div(b)?;

    // `scale()` is u32; the ideal can be negative (a coarser dividend than
    // divisor), which simply means "no padding required".
    let ideal_scale = i64::from(a.scale()) - i64::from(b.scale());

    // Minimal form first: this is what removes the EXTRA trailing zero in
    // `7 / 2 -> 3.50`. `normalize` never loses value.
    let mut result = quotient.normalize();
    let target = ideal_scale.max(i64::from(result.scale()));

    // `rescale` takes u32 and is a no-op past the mantissa's capacity; the
    // clamp keeps a pathological ideal from wrapping on the cast.
    if let Ok(target) = u32::try_from(target)
        && result.scale() < target
    {
        result.rescale(target);
    }
    Some(result)
}

/// Negate with Python `decimal`'s sign rule for zero.
///
/// Python defines unary minus as `0 - x`, so negating ANY zero yields a
/// POSITIVE zero:
///
/// ```text
///   -Decimal("0.00")   ==  0.00
///   -Decimal("-0.00")  ==  0.00
/// ```
///
/// `rust_decimal`'s `Neg` flips the sign bit unconditionally, so `-dec!(0.00)`
/// is a signed zero that renders `-0.00`. beancount normalizes it away — a
/// ledger posting written `-0.00 CNY` loads as `Decimal('0.00')` there, and
/// bean-query prints `0.00` — while rledger's parser applies the sign with a
/// bare `-n` and kept the `-0.00` all the way to the output.
///
/// A zero has no sign in bookkeeping; this is the canonical negation for any
/// site that flips a parsed or computed amount.
#[must_use]
pub fn negate_python(number: Decimal) -> Decimal {
    if number.is_zero() {
        // `abs` clears the sign bit and keeps the scale — `normalize` would
        // strip the scale too, and `+ ZERO` collapses it to `0`.
        number.abs()
    } else {
        -number
    }
}

/// Round to `dp` decimal places with Python `decimal`'s sign rule for zero.
///
/// Python's `quantize` keeps the sign when a small negative rounds away:
///
/// ```text
///   Decimal("-0.00495").quantize(Decimal("0.01"))  ==  -0.00
///   Decimal("-0.0000001").quantize(Decimal("0.01")) == -0.00
/// ```
///
/// `rust_decimal`'s `round_dp` returns an UNSIGNED zero there, which loses
/// the only information the cell still carries — that the underlying balance
/// is negative rather than exactly zero. bean-query renders `-0.00 USD` for a
/// `-0.00495 USD` position; rledger rendered `0.00 USD` and read as flat.
///
/// The result is padded to exactly `dp` (see the caller's note on `round_dp`
/// only ever reducing the scale).
#[must_use]
pub fn round_dp_python(number: Decimal, dp: u32) -> Decimal {
    let mut rounded = number.round_dp(dp);
    rounded.rescale(dp);
    if rounded.is_zero() && number.is_sign_negative() {
        // Re-apply the sign the rounding dropped. Negating a positive zero is
        // exactly how a signed zero is constructed in `rust_decimal`.
        return -rounded;
    }
    rounded
}

#[cfg(test)]
mod tests {
    use super::*;
    use rust_decimal_macros::dec;
    use std::str::FromStr;

    /// The exact shape `rust_decimal` gets wrong: a zero operand's scale is
    /// dropped. Both orders, since its zero shortcut applies to either side.
    ///
    /// Asserts on `to_string()`, NOT on `Decimal` equality. `==` compares
    /// value and ignores scale — `dec!(1) == dec!(1.00)` — so an
    /// `assert_eq!(add_python_scale(..), dec!(1.00))` here would pass against
    /// a plain `a + b` and pin nothing. That is exactly what this test looked
    /// like when Copilot caught it on #2046; the rendered form is the whole
    /// subject of the divergence, so it is what gets asserted.
    #[test]
    fn a_zero_operand_keeps_its_scale() {
        assert_eq!(add_python_scale(dec!(0.00), dec!(1)).to_string(), "1.00");
        assert_eq!(add_python_scale(dec!(1), dec!(0.00)).to_string(), "1.00");
        assert_eq!(sub_python_scale(dec!(0.00), dec!(1)).to_string(), "-1.00");
        assert_eq!(sub_python_scale(dec!(1), dec!(0.00)).to_string(), "1.00");
    }

    /// The trap the test above avoids, pinned so it cannot silently return:
    /// `Decimal`'s `==` cannot see a scale difference, and the raw operator
    /// really does drop it.
    #[test]
    fn decimal_equality_cannot_see_the_bug_but_rendering_can() {
        assert_eq!(dec!(1), dec!(1.00), "== ignores scale");
        assert_eq!((dec!(0.00) + dec!(1)).to_string(), "1", "the bug itself");
        assert_eq!(add_python_scale(dec!(0.00), dec!(1)).to_string(), "1.00");
    }

    /// Ordinary operands already behaved; the fix must not disturb them.
    #[test]
    fn non_zero_operands_are_unchanged() {
        for (a, b, want) in [
            (dec!(2.00), dec!(1), "3.00"),
            (dec!(1), dec!(2.00), "3.00"),
            (dec!(2.50), dec!(2.50), "5.00"),
            (dec!(111.11), dec!(-111.11), "0.00"),
            (dec!(1), dec!(2), "3"),
        ] {
            assert_eq!(add_python_scale(a, b).to_string(), want, "{a} + {b}");
        }
    }

    /// Padding is upward only — a result that already carries more scale than
    /// either operand (impossible for `+`, but the guard is what makes that
    /// true) keeps it, and no significant digit is ever dropped.
    #[test]
    fn never_truncates() {
        assert_eq!(add_python_scale(dec!(0.5), dec!(0.25)).to_string(), "0.75");
        assert_eq!(
            add_python_scale(dec!(0), dec!(1.2345)).to_string(),
            "1.2345"
        );
    }

    /// The running-total shape from the compat corpus: passing through zero
    /// mid-accumulation must not reset the scale for the terms that follow.
    #[test]
    fn accumulating_through_zero_keeps_the_widest_scale() {
        let terms = [
            dec!(-1),
            dec!(1),
            dec!(-111.11),
            dec!(111.11),
            dec!(-2),
            dec!(2),
        ];

        let mut python_like = Decimal::ZERO;
        let mut naive = Decimal::ZERO;
        for t in terms {
            python_like = add_python_scale(python_like, t);
            naive += t;
        }

        assert_eq!(python_like.to_string(), "0.00");
        assert_eq!(naive.to_string(), "0", "pins the pre-fix behavior");
    }

    /// Division scale, against Python `decimal`'s answers verbatim.
    ///
    /// Every expectation below was produced by running the case through
    /// `CPython`'s `decimal` (3.13) rather than reasoned out — the ideal-exponent
    /// rule is easy to state and easy to get subtly wrong, and a table of
    /// hand-derived expectations would just encode the same misreading twice.
    ///
    /// Asserts on `to_string()`: `Decimal`'s `==` ignores scale, which is the
    /// entire subject here.
    #[test]
    fn division_matches_python_decimal_scale() {
        // (dividend, divisor, python's rendering)
        let cases = [
            // Zero dividend — `rust_decimal` drops the scale (gives `0`).
            ("0.00", "4", "0.00"),
            ("0.000", "3", "0.000"),
            ("0.0", "7", "0.0"),
            ("0.00", "2.0", "0.0"),
            ("0.00", "1", "0.00"),
            ("-0.00", "3", "0.00"), // see the sign note below
            // Zero with no scale to keep.
            ("0", "4", "0"),
            // Exact quotients — `rust_decimal` OVER-pads `7 / 2` to `3.50`.
            ("7", "2", "3.5"),
            ("5", "4", "1.25"),
            ("1.0", "2.00", "0.5"),
            // Exact quotients both already agree on.
            ("1.00", "2", "0.50"),
            ("3.00", "3", "1.00"),
            ("1.000", "8", "0.125"),
            ("10.00", "4", "2.50"),
            ("2.50", "5", "0.50"),
            ("100.00", "8", "12.50"),
            ("12.345", "5", "2.469"),
            // Inexact: no trailing zeros to strip, scale far past the ideal,
            // so the rule leaves `rust_decimal`'s result alone.
            ("1", "3", "0.3333333333333333333333333333"),
        ];

        for (a, b, want) in cases {
            let a = Decimal::from_str(a).expect("dividend parses");
            let b = Decimal::from_str(b).expect("divisor parses");
            let got = checked_div_python_scale(a, b).expect("no overflow");
            assert_eq!(got.to_string(), want, "{a} / {b}");
        }
    }

    /// `from_str` drops a literal minus on zero, but the TYPE can hold a
    /// signed zero — the two are different things, and #2049 conflated them.
    ///
    /// That commit asserted "the type has no signed zero" from this same
    /// `from_str` evidence. It does: `-dec!(0.00)` renders `-0.00`, which is
    /// exactly the value the parser used to archive for a `-0.00` literal.
    /// The narrower true statement is pinned here instead, since the division
    /// table above depends on it: a `-0.00` DIVIDEND reaching
    /// `checked_div_python_scale` via `from_str` is already unsigned, so the
    /// `-0.00` Python would print is unreachable by that route.
    #[test]
    fn from_str_drops_the_sign_on_zero_though_the_type_can_carry_one() {
        let parsed = Decimal::from_str("-0.00").expect("parses");
        assert_eq!(parsed.to_string(), "0.00", "from_str drops it");
        assert!(!parsed.is_sign_negative());

        // ...but the type carries one when constructed by negation.
        assert_eq!((-Decimal::from_str("0.00").unwrap()).to_string(), "-0.00");

        assert_eq!(
            checked_div_python_scale(parsed, Decimal::from(3))
                .expect("no overflow")
                .to_string(),
            "0.00",
        );
    }

    /// Divide-by-zero is `None`, not a panic — BQL renders it NULL.
    #[test]
    fn division_by_zero_is_none() {
        assert_eq!(
            checked_div_python_scale(dec!(1.00), Decimal::ZERO),
            None,
            "div-by-zero must not panic",
        );
    }

    /// Negating a zero must yield a POSITIVE zero, as Python does.
    ///
    /// `rust_decimal`'s `Neg` flips the sign bit unconditionally, so a bare
    /// `-dec!(0.00)` renders `-0.00`. beancount loads a ledger posting written
    /// `-0.00 CNY` as `Decimal('0.00')` and bean-query prints `0.00`, so the
    /// signed zero was ours alone.
    ///
    /// Asserts on `to_string()` — `==` cannot see a sign on zero any more than
    /// it can see scale (`dec!(0.00) == -dec!(0.00)`), so a value-level
    /// assertion would pass against the bug.
    #[test]
    fn negating_a_zero_gives_an_unsigned_zero() {
        assert_eq!((-dec!(0.00)).to_string(), "-0.00", "the bug itself");
        assert_eq!(negate_python(dec!(0.00)).to_string(), "0.00");
        assert_eq!(negate_python(-dec!(0.00)).to_string(), "0.00");
        // Scale survives — `normalize()` would have flattened it to `0`.
        assert_eq!(negate_python(dec!(0.0000)).to_string(), "0.0000");
        // Non-zero negation is untouched.
        assert_eq!(negate_python(dec!(1.25)).to_string(), "-1.25");
        assert_eq!(negate_python(dec!(-1.25)).to_string(), "1.25");
    }

    /// Rounding a small negative to zero must KEEP the sign, as Python's
    /// `quantize` does — the opposite direction from the negation rule above,
    /// which is why one fix could not serve both.
    ///
    /// Expectations taken from `CPython`: `Decimal("-0.00495").quantize(
    /// Decimal("0.01"))` is `-0.00`. bean-query renders `-0.00 USD` for such a
    /// position; rledger rendered `0.00 USD`, which reads as an exactly flat
    /// balance when it is not.
    #[test]
    fn rounding_a_small_negative_to_zero_keeps_the_sign() {
        assert_eq!(dec!(-0.00495).round_dp(2).to_string(), "0.00", "the bug");

        for (value, dp, want) in [
            (dec!(-0.00495), 2, "-0.00"),
            (dec!(-0.004), 2, "-0.00"),
            (dec!(-0.0000001), 2, "-0.00"),
            (dec!(0.00495), 2, "0.00"),
            (dec!(-1.004), 2, "-1.00"),
            (dec!(-0.00495), 5, "-0.00495"),
            (dec!(1), 2, "1.00"),
        ] {
            assert_eq!(
                round_dp_python(value, dp).to_string(),
                want,
                "{value} at {dp}dp",
            );
        }
    }

    /// The checked variants exist so BQL can map a value-range overflow to
    /// NULL instead of panicking (`rust_decimal` panics on raw `+`). That
    /// path had no test — the one behavior the checked form is FOR.
    #[test]
    fn checked_variants_return_none_on_overflow() {
        assert_eq!(checked_add_python_scale(Decimal::MAX, Decimal::MAX), None);
        assert_eq!(checked_sub_python_scale(Decimal::MIN, Decimal::MAX), None);

        // And still compute the ordinary case, with the scale rule applied.
        assert_eq!(
            checked_add_python_scale(dec!(0.00), dec!(1))
                .expect("no overflow")
                .to_string(),
            "1.00"
        );
        assert_eq!(
            checked_sub_python_scale(dec!(1), dec!(0.00))
                .expect("no overflow")
                .to_string(),
            "1.00"
        );
    }

    /// The rescale step must never corrupt a value it cannot widen.
    ///
    /// A sum near `Decimal::MAX` has no room for extra fractional digits, so
    /// the requested scale is unreachable. `Decimal::rescale` is documented
    /// to leave the value alone in that case rather than truncating, and this
    /// pins that we depend on it: if it ever became saturating or truncating,
    /// the failure mode is a silently WRONG money value rather than a panic
    /// or an error, which nothing else here would catch.
    ///
    /// Verified against the real type rather than assumed — `MAX.rescale(2)`
    /// is a no-op returning `79228162514264337593543950335`.
    #[test]
    fn rescale_beyond_capacity_preserves_the_value() {
        let near_max = Decimal::MAX - Decimal::ONE;
        let sum = add_python_scale(near_max, dec!(0.00));

        assert_eq!(
            sum, near_max,
            "a value too large to carry the target scale must keep its VALUE",
        );
        assert_eq!(sum.to_string(), near_max.to_string());
    }
}