rug

• Bug fix: incorrect underflow when dividing Rational by [Float][flo-1-28] (issue 85).

• The crate now requires rustc version 1.85.0 or later.
  • The following methods were added to Float:
    • as_shl, as_shr
  • The following methods were added to Complex:
    • as_shl, as_shr
  • The experimental feature num-complex was added (issue 84).
  • The experimental feature borsh was added (merge request 6).

• The [az crate], which is a dependency, is now re-exported (issue 73).
  • The following methods were added to Float:
    • to_integer_in_place, to_integer_round_in_place, to_integer_exp_in_place (issue 75)
    • to_rational_in_place (issue 75)
    • remainder_quo31, remainder_quo31_mut, remainder_quo31_round (issue 76)
    • remainder_quo31_from, remainder_quo31_from_round
    • remainder_quo31_ref
    • frexp, frexp_mut, frexp_ref (issue 76)
  • Bug fix: <code>Float::to_f16</code> and <code>Float::to_f128</code> were giving bad results for some subnormal numbers (issue 80).

Compatibility note

• These subnormalization methods for Float now handle numbers with exponents smaller than subnormals by rounding them to either 0 or a subnormal with the minimum nonzero magnitude (issue 78):
  • subnormalize_ieee, subnormalize_ieee_round
  • subnormalize, subnormalize_round

• Bug fix: implementations of CompleteRound were wrongly using prec_t instead of [u32] for <code>CompleteRound::Prec</code> (issue 72).

• Bug fix: Complete was not implemented but documented for incomplete values produced by the following methods (issue 69): * <code>Integer::clamp_ref</code>, <code>Integer::invert_ref</code>, <code>Integer::pow_mod_ref</code> * <code>Integer::random_bits</code>, <code>Integer::random_below_ref</code>
  • <code>Float::get_significand</code> is now usable in const context.
  • The experimental feature nightly-float was added (issue 68).

• Bug fix: implementations of CompleteRound were wrongly using prec_t instead of [u32] for <code>CompleteRound::Prec</code> (issue 72).

• The following functions were added:
  • <code>float::prec_min_64</code>, <code>float::prec_max_64</code>
  • The <code>complex::Prec64</code> trait was added.
  • The following methods were added:
    • <code>Float::new_64</code>, <code>Float::with_val_64</code>, <code>Float::with_val_round_64</code>
    • <code>Float::prec_64</code>, <code>Float::set_prec_64</code>, <code>Float::set_prec_round_64</code>
    • <code>Complex::new_64</code>, <code>Complex::with_val_64</code>, <code>Complex::with_val_round_64</code>
    • <code>Complex::prec_64</code>, <code>Complex::set_prec_64</code>, <code>Complex::set_prec_round_64</code>

• Bug fix: a precondition for slice::from_raw_parts was being violated (issue 63).

• Float now implements <code>AssignRound<[bool][bool]></code>.
  • The following methods are now usable in const context:
    • <code>Float::as_neg</code>, <code>Float::as_abs</code>
    • <code>Complex::as_neg</code>, <code>Complex::as_conj</code>, <code>Complex::as_mul_i</code>
    • <code>MiniInteger::borrow</code>
    • <code>MiniRational::borrow</code>
    • <code>MiniFloat::borrow</code>
    • <code>MiniComplex::borrow</code>
  • The std optional feature was added (issue 62). The feature is enabled by default, and the crate will be compiled without the standard library (no_std) if the feature is disabled.
  • MiniRational now implements
    • <code>[From][From]<MiniInteger></code>, <code>Assign<MiniInteger></code>
  • MiniFloat now implements
    • <code>[From][From]<[bool][bool]></code>, <code>Assign<[bool][bool]></code>
  • MiniComplex now implements
    • <code>[From][From]<MiniFloat></code>, <code>Assign<MiniFloat></code>
    • <code>[From][From]<[(][tuple]MiniFloat[,][tuple] MiniFloat[)][tuple]></code>, <code>Assign<[(][tuple]MiniFloat[,][tuple] MiniFloat[)][tuple]></code>
  • The following methods were added to MiniInteger:
    • const_from_bool
    • const_from_i8, const_from_i16, const_from_i32, const_from_i64, const_from_i128, const_from_isize
    • const_from_u8, const_from_u16, const_from_u32, const_from_u64, const_from_u128, const_from_usize
  • The following method was added to MiniRational:
    • const_from_integer
  • The following methods were added to MiniFloat:
    • const_from_bool
    • const_from_i8, const_from_i16, const_from_i32, const_from_i64, const_from_i128, const_from_isize
    • const_from_u8, const_from_u16, const_from_u32, const_from_u64, const_from_u128, const_from_usize
    • const_from_f32, const_from_f64
    • const_from_special
  • The following methods were added to MiniComplex:
    • const_from_real, const_from_parts
  • When the num-traits experimental feature is enabled, the following traits are implemented for Integer:
    • CheckedDiv, CheckedRem
    • Euclid, CheckedEuclid

Compatibility notes

• The following methods no longer set the MPFR NaN flag if a NaN is encountered:
  • <code>Float::as_neg</code>, <code>Float::as_abs</code>
  • <code>Complex::as_neg</code>, <code>Complex::as_conj</code>, <code>Complex::as_mul_i</code>
• Functionality that depends on std now requires the std feature to be enabled. While the std feature is enabled by default, it is now disabled if the crate’s default features are disabled in [Cargo.toml]. In this case, std must be added to the list of features to reenable the functionality.
• The implementation of the deprecated SmallInteger, SmallRational, SmallFloat and SmallComplex was changed to remove [their soundness issue][issue 52]. They now allocate memory for the digits, though their whole point was to be non-allocating.

• MiniInteger, MiniRational, MiniFloat and MiniComplex were added to replace SmallInteger, SmallRational, SmallFloat and SmallComplex, which are now deprecated. * The new structs do not implement [Deref] directly, as this causes issues with the current Rust memory models (issue 52). * Borrowing requires the use of the borrow methods, which return an object that implements [Deref]. * The new borrow_excl methods were also added; they require exclusive access in order to save on some housekeeping and to return references directly without the need of [Deref]. * The new structs implement [Copy] and [Sync].

• Bug fix: implementations of [PartialOrd] and [PartialEq] between Rational numbers and tuples of two primitive integers were breaking the transitivity property of the traits, so the implementations were removed (issue 58). See the compatibility note below.
  • Direct [PartialOrd] and [PartialEq] comparisons are now implemented
    • between Integer and SmallInteger,
    • between Rational and SmallRational, and
    • between Float and SmallFloat.
  • Direct [PartialEq] comparisons are now implemented between Complex and SmallComplex.
  • SmallInteger can now be assigned or converted directly to Integer using Assign and [From].
  • SmallRational can now be assigned or converted directly to Rational using Assign and [From].
  • SmallFloat can now be assigned directly to Float using AssignRound and Assign.
  • SmallComplex can now be assigned directly to Complex using AssignRound and Assign.
  • [Debug] is now implemented for SmallInteger, SmallRational, SmallFloat and SmallComplex.
  • For the Complex struct, the method eq0 was renamed to is_zero. The old method name is deprecated.

Compatibility note

The implementations of [PartialOrd] and [PartialEq] between Rational numbers and tuples of two primitive integers were breaking the transitivity property of the two traits in question. Since the implementations necessarily break the trait guarantees, having the traits implemented is considered a bug (see issue 58 for more details). These buggy implementations were removed. To fix code that used these comparisons without adding extra allocations, SmallRational can be used. For example rational == (1, 3) can be replaced with rational == SmallRational::from((1, 3)).

• The following methods were added to Integer: * is_zero, is_positive, is_negative * modulo, modulo_mut, modulo_from, modulo_ref
  • The following methods were added to Rational:
    • is_zero, is_positive, is_negative
  • Bug fix: panic instead of raising floating-point exception when dividing by Rational zero (issue 53).
  • Complete was implemented for incomplete values produced by DivRounding and RemRounding implementations (merge request 4).

• Bug fix: <code>[Float][flo-1-20]::[to_f32_round][flo-tfr-1-20]</code> was rounding some numbers in the subnormal range incorrectly ([issue 56]).

• The [gmp-mpfr-sys] dependency was updated to version 1.6.
  • The following methods were added to Integer:
    • prev_prime, prev_prime_mut, prev_prime_ref
  • The following associated constants were added to Integer:
    • ONE, NEG_ONE
  • The following associated constants were added to Rational:
    • ZERO, ONE, NEG_ONE

• Bug fix: a zero denominator could be left after catching a panic from <code>[Rational][rat-1-19]::[mutate_numer_denom][rat-mnd-1-19]</code> ([issue 49]).

• Bug fix: <code>[Rational][rat-1-19]::[mutate_numer_denom][rat-mnd-1-19]</code> and <code>[Complex][com-1-19]::[mutate_real_imag][com-mri-1-19]</code> were not unwind-safe ([issue 47]).

• The crate now requires rustc version 1.65.0 or later.
  • The [gmp-mpfr-sys] dependency was updated to version 1.5.
  • The <code>Round::AwayZero</code> rounding mode was added.
  • The following methods were added to Float:
    • sin_u, sin_u_mut, sin_u_round, sin_u_ref
    • cos_u, cos_u_mut, cos_u_round, cos_u_ref
    • tan_u, tan_u_mut, tan_u_round, tan_u_ref
    • sin_pi, sin_pi_mut, sin_pi_round, sin_pi_ref
    • cos_pi, cos_pi_mut, cos_pi_round, cos_pi_ref
    • tan_pi, tan_pi_mut, tan_pi_round, tan_pi_ref
    • asin_u, asin_u_mut, asin_u_round, asin_u_ref
    • acos_u, acos_u_mut, acos_u_round, acos_u_ref
    • atan_u, atan_u_mut, atan_u_round, atan_u_ref
    • atan2_u, atan2_u_mut, atan2_u_round, atan2_u_ref
    • asin_pi, asin_pi_mut, asin_pi_round, asin_pi_ref
    • acos_pi, acos_pi_mut, acos_pi_round, acos_pi_ref
    • atan_pi, atan_pi_mut, atan_pi_round, atan_pi_ref
    • atan2_pi, atan2_pi_mut, atan2_pi_round, atan2_pi_ref
    • log2_1p, log2_1p_mut, log2_1p_round, log2_1p_ref
    • log10_1p, log10_1p_mut, log10_1p_round, log10_1p_ref
    • exp2_m1, exp2_m1_mut, exp2_m1_round, exp2_m1_ref
    • exp10_m1, exp10_m1_mut, exp10_m1_round, exp10_m1_ref
    • compound_i, compound_i_mut, compound_i_round, compound_i_ref
    • root_i, root_i_mut, root_i_round, root_i_ref
  • The following methods were added to Complex:
    • agm, agm_mut, agm_round, agm_ref
  • The following methods are now const functions:
    • <code>Integer::as_limbs</code>
    • <code>Integer::as_neg</code>, <code>Integer::as_abs</code>
    • <code>Integer::is_even</code>, <code>Integer::is_odd</code>
    • <code>Integer::cmp0</code>
    • <code>Rational::into_raw</code>, <code>Rational::as_raw</code>
    • <code>Rational::numer</code>, <code>Rational::denom</code>
    • <code>Rational::into_numer_denom</code>
    • <code>Rational::as_neg</code>, <code>Rational::as_abs</code>, <code>Rational::as_recip</code>
    • <code>Rational::cmp0</code>
    • <code>Rational::is_integer</code>
    • <code>Float::prec</code>
    • <code>Float::from_raw</code>, <code>Float::into_raw</code>, <code>Float::as_raw</code>
    • <code>Float::as_ord</code>, <code>Float::as_complex</code>
    • <code>Float::is_nan</code>, <code>Float::is_infinite</code>, <code>Float::is_finite</code>, <code>Float::is_zero</code>, <code>Float::is_normal</code>, <code>Float::classify</code>
    • <code>Float::cmp0</code>
    • <code>Float::get_exp</code>
    • <code>Float::is_sign_positive</code>, <code>Float::is_sign_negative</code>
    • <code>Complex::prec</code>
    • <code>Complex::from_raw</code>, <code>Complex::into_raw</code>, <code>Complex::as_raw</code>
    • <code>Complex::real</code>, <code>Complex::imag</code>
    • <code>Complex::into_real_imag</code>, <code>Complex::borrow_real_imag</code>
    • <code>Complex::as_ord</code>
    • <code>Complex::eq0</code>
  • BorrowInteger, BorrowRational, BorrowFloat and BorrowComplex now implement [Clone], [Copy] and formatting traits; and each have a static method const_deref.

• For the Integer struct, the methods gcd_cofactors, gcd_cofactors_mut and gcd_cofactors_ref were renamed to extended_gcd, extended_gcd_mut and extended_gcd_ref. The old method names are deprecated.
  • All error types now implement [Clone], [Copy], [PartialEq] and [Eq].
  • The IntegerExt64 extension trait was added to support very large integers better on some 64-bit platforms.

• The new method <code>Integer::assign_bytes_radix_unchecked</code> was added (issue 41).
  • Integer now implements [LowerExp] and [UpperExp] for convenience by transparently converting the Integer to a Float (issue 42).
  • Bug fix: the CompleteRound trait is now implemented for
    • the value returned from <code>Float::mul_add_mul_ref</code>
    • the value returned from <code>Float::mul_sub_mul_ref</code>
  • The CompleteRound trait is now also implemented for
    • the value returned from negating a reference to Float, that is <code><&Float as [Neg][Neg]>::[Output][NegOutput]</code>
    • the value returned from negating a reference to Complex, that is <code><&Complex as [Neg][Neg]>::[Output][NegOutput]</code>

• Now Float implements Cast, CheckedCast and UnwrappedCast with primitive integers as the destination types, in addition to the previously supported SaturatingCast.
  • The new methods <code>Float::total_cmp</code> and <code>Complex::total_cmp</code> were added.
  • The new methods <code>Complex::borrow_real_imag</code> and <code>Complex::mutate_real_imag</code> were added.
  • Arithmetic operators are overloaded to work on combinations of Rug types and [isize] and [usize].
  • The BorrowInteger, BorrowRational, BorrowFloat and BorrowComplex structs were added.

• Now Integer, Rational, Float and Complex implement [Shl], [Shr], [ShlAssign] and [ShrAssign] with [isize] and [usize] as the type of the right-hand side, in addition to the previously supported [i32] and [u32].
  • The new experimental feature num-traits was added to implement some traits from the [num-traits crate] and the [num-integer crate] (issue 30).

• Bug fix: <code>[Rational][rat-1-14]::[to_f32][rat-tf-1-14]</code> was rounding away from zero in some cases when the rational number should be converted to a subnormal number ([issue 36]).

• Bug fix: OrdFloat was incorrectly ordering +NaN < −∞.
  • The <code>Integer::ZERO</code> associated constant was added.
  • The <code>Integer::shrink_to</code> method was added.
  • The <code>Rational::is_integer</code> method was added.
  • The <code>Complete::complete_into</code> provided method was added to the Complete trait.
  • The CompleteRound trait was added to make it easier to convert Float and Complex numbers incomplete-computation values to their final value.

• The incomplete-computation value returned by <code>Integer::{sum,dot}</code> can now be subtracted from an Integer.
  • The incomplete-computation value returned by <code>Rational::{sum,dot}</code> can now be subtracted from a Rational number.
  • The incomplete-computation value returned by <code>Float::{sum,dot}</code> can now be subtracted from a Float.
  • The incomplete-computation value returned by <code>Complex::{sum,dot}</code> can now be subtracted from a Complex number.
  • The incomplete-computation value returned by <code>Integer::square_ref</code> can now be added to/subtracted from an Integer.
  • The Complete trait was added to make it easier to convert incomplete-computation values to their final value.
  • The <code>Round::reverse</code> method was added.

• A new method <code>Integer::as_limbs</code> was added.
  • The [az crate] dependency was updated to version 1.1.

• The [gmp-mpfr-sys] dependency was updated to version 1.4.
  • Now it is possible to display Float numbers with only one siginificant digit.

• The [gmp-mpfr-sys] dependency was updated to version 1.3.
  • Now <code>[Option][Option]<Integer></code> has the same size as Integer; and similar for Rational, Float, Complex, RandState and ThreadRandState.

• Displaying Float and Complex numbers, and converting them to strings, now avoids outputting an exponent term if the radix point can be moved to the correct place without inserting any extra digits. For example "42.0" will be produced rather than "4.20e1" (issue 18). This change does not affect output when [LowerExp] ("{:e}" format) or [UpperExp] ("{:E}" format) is used.
  • New methods <code>Float::to_sign_string_exp</code> and <code>Float::to_sign_string_exp_round</code> were added.
  • A new function <code>float::allowed_exp_range</code> was added.
  • A new method <code>Float::clamp_exp</code> was added.
  • The following methods are now const functions:
    • <code>Integer::from_raw</code>, <code>Rational::from_raw</code>
  • The [az crate] is now a public dependency, and wrapping and checked casts to/from primitives and big numbers are provided through the traits of the crate.

Compatibility note

The output of Float and Complex numbers was changed as specified above.

• The following methods are now const functions: * <code>Integer::new</code>, <code>Integer::into_raw</code>, <code>Integer::as_raw</code> * <code>SmallInteger::new</code>, <code>SmallRational::new</code>
  • A new method <code>Integer::as_rational</code> was added.
  • A new method <code>Float::as_complex</code> was added.
  • SmallFloat and SmallComplex can now be initialized with Special.
  • New methods <code>SmallFloat::new</code> and <code>SmallComplex::new</code> were added.
  • SmallFloat and SmallComplex now implement [Default].
  • Integer now implements <code>[AsRef][AsRef]<[[][slice]limb_t[]][slice]></code>.
  • Float now implements <code>[AsRef][AsRef]<OrdFloat></code>, and OrdFloat now implements <code>[AsRef][AsRef]<Float></code> and <code>[AsMut][AsMut]<Float></code>.
  • Complex now implements <code>[AsRef][AsRef]<OrdComplex></code>, and OrdComplex now implements <code>[AsRef][AsRef]<Complex></code> and <code>[AsMut][AsMut]<Complex></code>.

• The crate now requires rustc version 1.37.0 or later.
  • The [gmp-mpfr-sys] dependency was updated to version 1.2.
  • The RemAssignRound and RemFromRound traits were added.
  • The [Rem], [RemAssign], RemFrom, RemAssignRound and RemFromRound traits were implemented for Float.
  • Arithmetic operations with one Integer or integer primitive operand and one Rational operand were added.
  • A new method <code>Integer::div_exact_from</code> was added.
  • New methods <code>Integer::gcd_u</code>, <code>Integer::gcd_u_mut</code> and <code>Integer::gcd_u_ref</code>were added.
  • New methods <code>Integer::lcm_u</code>, <code>Integer::lcm_u_mut</code> and <code>Integer::lcm_u_ref</code>were added.
  • New methods <code>Float::remainder</code>, <code>Float::remainder_mut</code>, <code>Float::remainder_round</code>, <code>Float::remainder_from</code>, <code>Float::remainder_from_round</code> and <code>Float::remainder_ref</code> were added.

Compatibility note

• SmallInteger, SmallRational, SmallFloat and SmallComplex are no longer [Sync] to avoid the possibility of a data race. References obtained by dereferencing them, for example the <code>&Integer</code> returned from <code>SmallInteger::[deref][deref]</code>, are still [Send] and [Sync].

• Arithmetic operator implementations for [i8], [i16], [i64], [i128], [u8], [u16], [u64] and [u128] were added to the existing implementations for [i32] and [u32].
  • A new function <code>float::free_cache</code> and its supporting enum FreeCache were added.
  • A new method <code>RandState::into_custom_boxed</code> was added.
  • A new struct ThreadRandState and a new trait ThreadRandGen were added to the rand module to enable the use of custom random generators that are not [Send] and [Sync].
  • Numeric type methods which take RandState now can also take ThreadRandState.

Compatibility note

• The numeric type methods which took <code>&mut RandState</code> were changed to take <code>&mut dyn MutRandState</code> instead. Under normal use, this does not have any affect apart from allowing the use of <code>&mut ThreadRandState</code> as well. With function inlining and optimization, the generated code should have the same performance.

• Bug fix: <code>[Pow][pow-1-5]<[i32][i32]> for [Rational][rat-1-5]</code> was returning the reciprocal of the correct result.

• Bug fix: a memory leak in conversions of [Float][flo-1-5] to string was fixed (https://gitlab.com/tspiteri/rug/-/work_items/11).

• A new method <code>Integer::assign_digits_unaligned</code> was added to enable reading digits from unaligned memory.
  • A new method <code>Integer::write_digits_unaligned</code> was added to enable writing digits to unaligned and uninitialized memory.
  • New methods <code>Float::u_exp</code> and <code>Float::i_exp</code> were added.
  • A new method <code>Complex::abs_round</code> was added.
  • The documentation examples on from_raw methods now use [MaybeUninit] instead of <code>[mem][mem]::[uninitialized][uninitialized]</code>.

• The crate now requires rustc version 1.31.0 or later.
  • The method <code>RandState::as_raw</code> was fixed to take &self instead of &mut self.
  • <code>[float][flom-1-0]::prec_min</code> and <code>[float][flom-1-0]::prec_max</code> are now const functions.
  • New methods <code>Float::copysign</code>, <code>Float::copysign_mut</code> and <code>Float::copysign_ref</code> were added.

• A new method <code>SmallRational::assign_canonical</code> was added.

• Fixed bug in <code>[Integer][int-1-2]::[assign_digits][int-ad-1-2]</code>. (Thanks: Eric Scrivner)

• The [NotAssign][noa-1-2], [BitAndFrom][baf-1-2], [BitOrFrom][bof-1-2] and [BitXorFrom][bxf-1-2] traits were implemented for [bool].
  • The [NegAssign][nea-1-2] trait was implemented for [f32] and [f64].

• The [Integer][int-1-2] methods [from_digits][int-fd-1-2], [assign_digits][int-ad-1-2], [significant_digits][int-sd-1-2], [to_digits][int-td-1-2] and [write_digits][int-wd-1-2] now support [bool] [slices][slice].

• The implementations of [Sum] and [Product] for Integer and Rational were generalized to accept more types of elements.
  • New methods <code>Integer::keep_signed_bits</code>, <code>Integer::keep_signed_bits_mut</code> and <code>Integer::keep_signed_bits_ref</code> were added.
  • New methods <code>Integer::sum</code>, <code>Integer::dot</code> and <code>Integer::product</code> were added
  • New methods <code>Rational::sum</code>, <code>Rational::dot</code> and <code>Rational::product</code> were added.
  • New methods <code>Float::dot</code> and <code>Complex::dot</code> were added.

• Support for unstable [i128], [u128] and [TryFrom] was added in nightly.

• Support for [i128] and [u128] conversions and comparisons was added, conditional on compiler support.
  • Conditional on compiler support, [TryFrom] conversions were implemented for conversions
    • from Integer values to integer primitives,
    • from floating-point primitives to Rational numbers, and
    • from Float values to Rational numbers.
  • A new method <code>Float::get_significand</code> was added.
  • New methods <code>Float::u_pow_u</code> and <code>Float::i_pow_u</code> were added.
  • New methods <code>Integer::from_digits</code>, <code>Integer::to_digits</code>, <code>Integer::assign_digits</code>, <code>Integer::write_digits</code> and <code>Integer::significant_digits</code> were added, providing reading from and writing to slices of unsigned integer primitives.

• A bug in <code>[Integer][int-1-0]::[reserve][int-r-1-0]</code>, which was shrinking the capacity in some cases, was fixed.
  • The [gmp-mpfr-sys] dependency in [Cargo.toml] was fixed to use tilde instead of caret, since Rug uses internal implementation details.

• A new method <code>[Integer][int-1-0]::[is_power_of_two][int-ipot-1-0]</code> was added.
  • A new method <code>[Integer][int-1-0]::[signed_bits][int-sb-1-0]</code> was added.
  • New methods <code>[Integer][int-1-0]::[secure_pow_mod][int-spm-1-0]</code>, <code>[Integer][int-1-0]::[secure_pow_mod_mut][int-spmm-1-0]</code> and <code>[Integer][int-1-0]::[secure_pow_mod_ref][int-spmr-1-0]</code> were added.
  • New methods <code>[Integer][int-1-0]::[div_rem_round][int-drr-1-0]</code>, <code>[Integer][int-1-0]::[div_rem_round_mut][int-drrm-1-0]</code> and <code>[Integer][int-1-0]::[div_rem_round_ref][int-drrr-1-0]</code> were added.
  • A new method <code>[Complex][com-1-0]::[eq0][com-e-1-0]</code> was added.
  • Documentation was improved.

• The methods <code>Integer::invert_mut</code> and <code>Integer::pow_mod_mut</code> were changed to return <code>[Result][Result]<(), ()></code> instead of [bool].
  • The <code>float::Round</code>, <code>float::Constant</code> and <code>float::Special</code> enums are now marked as non-exhaustive.
  • All deprecated items were removed.
  • Unsound blanket implementations constrained on <code>T where SmallInteger: Assign<T></code> inside SmallRational and on <code>T where SmallFloat: Assign<T></code> inside SmallComplex were removed.

• Integer::invert_ref and Integer::pow_mod_ref now return an Option, not an object that is assignable to Result.
  • Float::random_bits and Complex::random_bits are now assignable to the number values, not to Result objects.
  • Rational::signum, Rational::trunc, Rational::ceil, Rational::floor and Rational::round now return Rational.
  • Complex::abs, Complex::arg and Complex::norm now return Complex.
  • Remove Default implementations from all types of Float and Complex; now the precision always has to be specified.
  • Remove Sum and Product implementations for Float and Complex.
  • Remove Clone and Copy implementations from all incomplete computation types.
  • Revamp top-level crate documentation.
  • Add Integer::parse and Rational::parse, and deprecate ValidInteger, ValidRational, valid_str_radix methods, and assign_str* methods.
  • Add Float::parse and Complex::parse, and deprecate ValidFloat, ValidComplex, from_str* methods, valid_str_radix methods, and assign_str* methods.
  • Rename Integer::gcd_coeffs* methods to Integer::gcd_cofactors*.
  • Integer::gcd_cofactors_ref now supports computing only one cofactor.
  • Deprecate Rational::to_integer and Rational::as_numer_denom.
  • Deprecate Rational::as_mut_numer_denom and replace with Rational::mutate_numer_denom.
  • Deprecate Complex::as_real_imag.

• Add Integer::square and Rational::square methods.
  • Add Rational::cmp_abs method.
  • Add Float::sum and Complex::sum methods.
  • Add from_raw, into_raw, as_raw and as_raw_mut to RandState.
  • Add RandGen::boxed_clone and RandState::new_custom_boxed, and thus support for cloning custom random generators.
  • Fix Float::PartialOrd<f32> (and <f64>) to return None when the primitive is NaN.

• Require rustc version 1.18.0.
  • Require gmp-mpfr-sys version 1.1.0.
  • Deprecate most assign_* methods, and replace with static methods that return an assignable object.
  • Deprecate Rational::copy_to_integer method.
  • Add Rational::assign_canonical and Rational::from_canonical methods.
  • Add Float::ln_u method.
  • Add Float::round_even methods.
  • Add Float::gamma_inc methods.
  • Add Float::random_normal and Float::random_exp methods.
  • Deprecate Float::assign_random_gaussian methods.
  • Add Complex::cmp_abs method.
  • Add Complex::root_of_unity method.
  • Deprecate SmallRational::from_canonicalized_* methods and replace with SmallRational::from_canonical method.
  • Deprecate SmallRational::assign_canonicalized_* methods.
  • Add as_nonreallocating_* methods to SmallInteger, SmallRational, SmallFloat and SmallComplex.
  • Fix SmallFloat::new and SmallComplex::new to produce numbers with a precision of 53.
  • Deprecate and hide float::Round::AwayFromZero.
  • Add Integer::signum, Rational::signum and Float::signum methods.
  • Add missing conversion to/from and comparisons to primitive types.
  • Add from_raw, into_raw, as_raw and as_raw_mut to Integer, Rational, Float and Complex.
  • Add Float::classify method.
  • Add Float::mul_add, Float::mul_sub, Float::mul_add_mul and Float::mul_sub_mul methods.
  • Add Complex::mul_add and Complex::mul_sub methods.

• Implement mathematical operations where operands include references to primitives.
  • Remove undefined behaviour: replace mem::swap(&mut src, &mut uninitialized) with ptr::copy_nonoverlapping(&src, &mut uninitialized, 1).

• Move rug::float::AssignRound to rug::ops::AssignRound.
  • OrdFloat now orders +NaN above +∞, while −NaN is still below −∞.
  • Change Float::subnormalize methods to require explicit minimum normal exponent.
  • Add Float::subnormalize_ieee methods to deduce exponent range from precision, like old Float::subnormalize. The new method also supports all IEEE 754-2008 precisions corresponding to k storage bits where k ≥ 128 and k is a multiple of 32.
  • Deprecate Rational::fract methods and replace with Rational::rem_trunc methods.
  • Add Rational::rem_ceil and Rational::rem_floor methods.
  • Add Rational::rem_round and Rational::fract_round methods.
  • Add Float::next_toward, Float::next_up and Float::next_down.
  • Add optional serde support.

• Rename Integer::sign, Rational::sign and Float::sign methods as Integer::cmp0, Rational::cmp0 and Float::cmp0.
  • Rename Float::pos_diff as Float::positive_diff.
  • Move rug::AssignRound to rug::float::AssignRound.
  • Add Integer::clamp, Rational::clamp and Float::clamp methods.
  • Add Integer::div_rem_ceil, Integer::div_rem_floor and Integer::div_rem_euc methods.
  • Add DivRounding, DivRoundingAssign and DivRoundingFrom traits, and their Rem counterparts, and implement them for Integer, and for combinations of Integer with i32 or u32.
  • Add Rational::fract_floor and Rational::fract_ceil methods.

• Fix swapped Float::sin_cos_mut and Float::sin_cos_round, Float::sinh_cosh_mut and Float::sinh_cosh_round, and Float::trunc_fract_mut and Float::trunc_fract_round.
  • Fix Float::to_f32_round.
  • Now Float::subnormalize only works for precisions defined in IEEE 754.
  • Add Integer::gcd_coeffs methods.

• Requires rustc version 1.17.0.
  • Rename Float::abs_diff as Float::pos_diff.
  • Replace Float::get_sign with Float::is_sign_positive and Float::is_sign_negative.
  • Add various as_neg, as_abs and as_conj methods.
  • Add OrdFloat and OrdComplex for complete ordering.