* libguile/numbers.c (scm_min, scm_max): Properly order the real
infinities and NaNs, per R6RS, and also take care to handle signed
zeroes properly. Note that this ordering is different than that of
`<', `>', `<=', and `>=', which return #f if any argument is a real
NaN, and consider the real zeroes to be equal. The relevant real
infinity (-inf.0 for min, +inf.0 for max) beats everything, including
NaNs, and NaNs beat everything else. Previously these were handled
improperly in some cases, e.g.:
(min 1/2 +nan.0) now returns +nan.0 (previously returned 0.5),
(max 1/2 +nan.0) now returns +nan.0 (previously returned 0.5),
(min -inf.0 +nan.0) now returns -inf.0 (previously returned +nan.0),
(max +inf.0 +nan.0) now returns +inf.0 (previously returned +nan.0),
(min -0.0 0.0) now returns -0.0 (previously returned 0.0),
(max 0.0 -0.0) now returns 0.0 (previously returned -0.0),
(max 0 -0.0) now returns 0.0 (previously returned -0.0),
(max -0.0 0 ) now returns 0.0 (previously returned -0.0).
* test-suite/tests/numbers.test (min, max): Add many more test cases
relating to NaNs, infinities, and signed zeroes. Change most existing
test cases to use `eqv?' instead of `=', in order to check exactness.
* libguile/numbers.c (scm_product): Handle exact 0 differently. A
product containing an exact 0 now returns an exact 0 if and only if
the other arguments are all exact. An inexact zero is returned if and
only if the other arguments are all finite but not all exact. If an
infinite or NaN value is present, a NaN value is returned.
Previously, any product containing an exact 0 yielded an exact 0,
regardless of the other arguments.
A note on the rationale for (* 0 0.0) returning 0.0 and not exact 0:
The exactness propagation rules allow us to return an exact result in
the presence of inexact arguments only if the values of the inexact
arguments do not affect the result. In this case, the value of the
inexact argument _does_ affect the result, because an infinite or NaN
value causes the result to be a NaN.
A note on the rationale for (* 0 +inf.0) being a NaN and not exact 0:
The R6RS requires that (/ 0 0.0) return a NaN value, and that (/ 0.0)
return +inf.0. We would like (/ x y) to be the same as (* x (/ y)),
and in particular, for (/ 0 0.0) to be the same as (* 0 (/ 0.0)),
which reduces to (* 0 +inf.0). Therefore (* 0 +inf.0) should return
a NaN.
* test-suite/tests/numbers.test: Add many multiplication tests.
* NEWS: Add NEWS entry.
* test-suite/tests/numbers.test: (real-nan?, complex-nan?,
imaginary-nan?): Add more discriminating NaN testing predicates
internal to numbers.test, and convert several uses of `nan?'
to use these instead:
* `real-nan?' checks that its argument is real and a NaN.
* `complex-nan?' checks that both the real and imaginary
parts of its argument are NaNs.
* `imaginary-nan?' checks that its argument's real part
is zero and the imaginary part is a NaN.
* libguile/numbers.c (scm_rationalize): Fix bugs. Previously, it
returned exact integers unmodified, although that was incorrect if
the epsilon was at least 1 or inexact, e.g. (rationalize 4 1) should
return 3 per R5RS and R6RS, but previously it returned 4. Also
handle cases involving infinities and NaNs properly, per R6RS.
* test-suite/tests/numbers.test: Add test cases for `rationalize'.
* NEWS: Add NEWS entry
* libguile/numbers.c (scm_integer_expt): No longer require that the
first argument be a number, in order to improve extensibility. This
allows us to efficiently raise arbitrary objects to an integer power
as long as we can multiply those objects. For example, this allows us
to efficiently exponentiate matrices if we define only multiplication
methods for matrices. Note also that scm_expt calls this procedure
whenever the exponent is an integer, regardless of the type of the
first argument. Also rearrange the order in which we test special
cases.
* test-suite/tests/numbers.test (expt, integer-expt): Comment out tests
that required `(expt #t 0)' and `(integer-expt #t 0)' to throw
exceptions. Add tests for (expt #t 2) and `(integer-expt #t 2)
instead.
* NEWS: Add NEWS entry
* test-suite/tests/numbers.test (test-eqv?): Remove special handling of
zeroes. Zeroes are now compared like all other numbers. Exact
numbers are compared with `eqv?' and inexact numbers are compared to
within test-epsilon.
Rework the testing framework for number-theoretic division operators:
`euclidean/', `euclidean-quotient', `euclidean-remainder',
`centered/', `centered-quotient', and `centered-remainder'.
Previously we compared all test results against a simple scheme
implementation of the same operations. However, these operations have
discontinuous jumps where a tiny change in the inputs can lead to a
large change in the outputs, e.g.:
(euclidean/ 130.00000000000 10/7) ==> 91.0 and 0.0
(euclidean/ 129.99999999999 10/7) ==> 90.0 and 1.42857142856141
In the new testing scheme, we compare values against the simple
implementations only if the input arguments contain an infinity or a
NaN. In the common case of two finite arguments, we simply make sure
that the outputs of all three operators (e.g. `euclidean/',
`euclidean-quotient', `euclidean-remainder') equal each other, that
outputs are exact iff both inputs are exact, and that the required
properties of the operator are met: that Q is an integer, that R is
within the specified range, and that N = Q*D + R.
* libguile/numbers.c (scm_quotient, scm_remainder, scm_modulo,
scm_zero_p, scm_positive_p, scm_negative_p, scm_real_part,
scm_imag_part, scm_numerator, scm_denominator, scm_magnitude,
scm_angle, scm_exact_to_inexact): Change from SCM_GPROC to
SCM_PRIMITIVE_GENERIC. As a side effect, all of these procedures now
have documentation strings.
(scm_exact_p, scm_inexact_p, scm_odd_p, scm_even_p, scm_finite_p,
scm_inf_p, scm_nan_p, scm_expt, scm_inexact_to_exact, scm_log,
scm_log10, scm_exp, scm_sqrt): Change from SCM_DEFINE to
SCM_PRIMITIVE_GENERIC, and make sure the code allows these functions
to be extended in practice.
(scm_real_part, scm_imag_part, scm_numerator, scm_denominator,
scm_inexact_to_exact): Simplify type dispatch code.
(scm_sqrt): Rename formal argument from x to z, since complex numbers
are supported.
(scm_abs): Fix empty FUNC_NAME.
* libguile/numbers.h (scm_finite_p): Add missing prototype.
(scm_inf_p, scm_nan_p): Rename formal parameter from n to x, since
the domain is the real numbers.
* test-suite/tests/numbers.test: Test for documentation strings. Change
from `expect-fail' to `pass-if' for several of these, and add tests
for others. Also add other tests for `real-part' and `imag-part',
which previously had none.
* libguile/numbers.c (scm_euclidean_quo_and_rem, scm_euclidean_quotient,
scm_euclidean_remainder, scm_centered_quo_and_rem,
scm_centered_quotient, scm_centered_remainder): New extensible
procedures `euclidean/', `euclidean-quotient', `euclidean-remainder',
`centered/', `centered-quotient', `centered-remainder'.
* libguile/numbers.h: Add function prototypes.
* module/rnrs/base.scm: Remove incorrect stub implementations of `div',
`mod', `div-and-mod', `div0', `mod0', and `div0-and-mod0'. Instead do
renaming imports of `euclidean-quotient', `euclidean-remainder',
`euclidean/', `centered-quotient', `centered-remainder', and
`centered/', which are equivalent to the R6RS operators.
* module/rnrs/arithmetic/fixnums.scm (fxdiv, fxmod, fxdiv-and-mod,
fxdiv0, fxmod0, fxdiv0-and-mod0): Remove redundant checks for division
by zero and unnecessary complexity.
(fx+/carry): Remove unneeded calls to `inexact->exact'.
* module/rnrs/arithmetic/flonums.scm (fldiv, flmod, fldiv-and-mod,
fldiv0, flmod0, fldiv0-and-mod0): Remove redundant checks for division
by zero and unnecessary complexity. Remove unneeded calls to
`inexact->exact' and `exact->inexact'
* test-suite/tests/numbers.test: (test-eqv?): New internal predicate for
comparing numerical outputs with expected values.
Add extensive test code for `euclidean/', `euclidean-quotient',
`euclidean-remainder', `centered/', `centered-quotient',
`centered-remainder'.
* test-suite/tests/r6rs-arithmetic-fixnums.test: Fix some broken test
cases, and remove `unresolved' test markers for `fxdiv', `fxmod',
`fxdiv-and-mod', `fxdiv0', `fxmod0', and `fxdiv0-and-mod0'.
* test-suite/tests/r6rs-arithmetic-flonums.test: Remove `unresolved'
test markers for `fldiv', `flmod', `fldiv-and-mod', `fldiv0',
`flmod0', and `fldiv0-and-mod0'.
* doc/ref/api-data.texi (Arithmetic): Document `euclidean/',
`euclidean-quotient', `euclidean-remainder', `centered/',
`centered-quotient', and `centered-remainder'.
(Operations on Integer Values): Add cross-references to `euclidean/'
et al, from `quotient', `remainder', and `modulo'.
* doc/ref/r6rs.texi (rnrs base): Improve documentation for `div', `mod',
`div-and-mod', `div0', `mod0', and `div0-and-mod0'. Add
cross-references to `euclidean/' et al.
* NEWS: Add NEWS entry.
* libguile/numbers.c (scm_rational_p): Return #f for infinities and
NaNs, per R6RS. Previously it returned #t for real infinities
and NaNs. They are still considered real by scm_real `real?'
however, per R6RS. Also simplify the code.
(scm_real_p): New implementation to reflect the fact that the
rationals and reals are no longer the same set. Previously it just
called scm_rational_p.
(scm_integer_p): Simplify the code.
* test-suite/tests/numbers.test: Add test cases for `rational?'
and `real?' applied to infinities and NaNs.
* doc/ref/api-data.texi (Real and Rational Numbers): Update docs to
reflect the fact that infinities and NaNs are no longer rational, and
that `real?' no longer implies `rational?'. Improve discussion of
infinities and NaNs.
* NEWS: Add NEWS entries, and combine with an earlier entry about
infinities no longer being integers.
Change `equal?' to work like `eqv?' for numbers.
Previously they worked differently in some cases, e.g.
when comparing signed zeroes or NaNs. For example,
(equal? 0.0 -0.0) returned #t but (eqv? 0.0 -0.0)
returned #f, and (equal? +nan.0 +nan.0) returned #f
but (eqv? +nan.0 +nan.0) returned #t.
* libguile/numbers.c (scm_real_equalp, scm_bigequal,
scm_complex_equalp, scm_i_fraction_equalp): Move to eq.c.
* libguile/eq.c (scm_real_equalp): Compare flonums using
real_eqv instead of ==, so that NaNs are now considered
equal, and to distinguish signed zeroes.
(scm_complex_equalp): Compare real and imaginary
components using real_eqv instead of ==, so that NaNs are
now considered equal, and to distinguish signed zeroes.
(scm_bigequal): Use scm_i_bigcmp instead of duplicating it.
(real_eqv): Test for NaNs using isnan(x) instead of
(x != x), and use SCM_UNLIKELY for optimization.
(scm_eqv_p): Use scm_bigequal, scm_real_equalp,
scm_complex_equalp, and scm_i_fraction_equalp to compare
numbers, instead of inline code. Those predicates now do
what scm_eqv_p formerly did internally. Replace if
statements with switch statements, as is done in
scm_equal_p. Remove useless code to check equality of
fractions with different SCM_CELL_TYPEs; this was for a
tentative "lazy reduction bit" which was never developed.
(scm_eqv_p, scm_equal_p): Remove useless code to check
equality between inexact reals and non-real complex numbers
with zero imaginary part. Such numbers do not exist,
because the current code is careful to never create them.
* test-suite/tests/numbers.test: Add test cases for
`eqv?' and `equal?'. Change existing test case for
`(equal? +nan.0 +nan.0)' to expect #t instead of #f.
* NEWS: Add NEWS entries.
* libguile/numbers.c (scm_difference, scm_product):
Fix bugs when negating SCM_MOST_POSITIVE_FIXNUM+1,
aka -SCM_MOST_NEGATIVE_FIXNUM. Previously, these cases
failed to normalize the result to a fixnum, causing
`=', `eqv?' and `equal?' to fail, e.g.:
(= most-negative-fixnum (- 0 (- most-negative-fixnum)))
(= most-negative-fixnum (* -1 (- most-negative-fixnum)))
(= most-negative-fixnum (* (- most-negative-fixnum) -1))
* test-suite/test/numbers.test: Add test cases to detect
bugs when negating SCM_MOST_POSITIVE_FIXNUM+1 and
SCM_MOST_NEGATIVE_FIXNUM by various methods.
* libguile/numbers.c (scm_inf_p, scm_finite_p, scm_nan_p): The domain of
these functions is the real numbers. Error on other input.
* doc/ref/api-data.texi (Reals and Rationals): Update the documentation
accordingly.
* test-suite/tests/numbers.test ("finite?", "inf?"): Update tests.
* libguile/numbers.c (scm_exact_p): Optimize by making use of the
SCM_INEXACTP macro.
(scm_inexact_p): Move it next to scm_exact_p, and add else's.
* test-suite/tests/numbers.test: Add test cases for `exact?'
and `inexact?' applied to infinities and NaNs.
* libguile/numbers.c (scm_finite_p): Add new predicate `finite?' from
R6RS to guile core, which returns #t if and only if its argument is
neither infinite nor a NaN. Note that this is not the same as (not
(inf? x)) or (not (infinite? x)), since NaNs are neither finite nor
infinite.
* test-suite/tests/numbers.test: Add test cases for `finite?'.
* module/rnrs/base.scm: Import `inf?' as `infinite?' instead of
reimplementing it. Previously, the R6RS implementation of
`infinite?' did not detect non-real complex infinities, nor did it
throw exceptions for non-numbers. (Note that NaNs _are_ considered
numbers by scheme, despite their name).
Import `finite?' instead of reimplementing it. Previously, the R6RS
implementation of `finite?' returned #t for both NaNs and non-real
complex infinities, in violation of R6RS.
* NEWS: Add NEWS entries, and reorganize existing numerics-related
entries together under one subheading.
* doc/ref/api-data.texi (Real and Rational Numbers): Add docs for
`finite?' and scm_finite_p.
* libguile/numbers.c (scm_is_integer): Infinities are not integers, per
the R6RS.
(scm_even_p, scm_odd_p): Passing an infinity to even? or odd? is an
error.
* test-suite/tests/numbers.test ("integer?"): Adapt test.
("expt"): Add tests for +inf.0 and -inf.0 exponents.
* NEWS: Add NEWS entries.
* libguile/numbers.c (scm_expt): Fix bug that caused expt to throw an
exception whenever the base was exact and the exponent was an
inexact integer, e.g. (expt 5 6.0).
(scm_expt): Fix bug that caused expt to introduce spurious imaginary
parts in the result when the base was an inexact negative real and
the exponent was an integer, e.g. (expt -1.0 2)
(scm_integer_expt, scm_expt): Change behavior of (integer-expt 0 -1),
and therefore also (expt 0 -1), to return NaN, per R6RS (actually,
R6RS says we should throw an exception or return an "unspecified
number object", but for now we use NaN). Formerly we returned 0, per
R5RS. R5RS claims that 0^x=0 for all non-zero x, but that's
mathematically incorrect, and probably an oversight.
(scm_integer_expt): Consistently throw a wrong-argument-type exception
when the exponent is inexact. Formerly, it didn't always check this
if the base was 0, 1, or -1.
* test-suite/tests/numbers.test ("integer-expt", "expt"): Add tests.
This results in a 17% improvement in the execution time of the "+" and
"-" benchmarks for fixnums.
* libguile/vm-i-scheme.c (ASM_ADD, ASM_SUB)[defined __x86_64__ &&
SCM_GNUC_PREREQ (4, 5)]: New macros.
(add)[defined ASM_ADD]: Use `ASM_ADD' for the fast path.
(sub)[defined ASM_SUB]: Use `ASM_SUB' for the fast path.
* test-suite/tests/numbers.test ("+")["fixnum + fixnum = bignum
(32-bit)", "fixnum + fixnum = bignum (64-bit)", "bignum + fixnum =
fixnum", "wrong type"]: New tests.
("-")["fixnum - fixnum = bignum (32-bit)", "fixnum - fixnum = bignum
(64-bit)", "bignum - fixnum = fixnum", "wrong type"]: New tests.
* test-suite/tests/00-initial-env.test ("goopsless")["+ wrong type
argument"]: Use `with-test-prefix/c&e' instead of `with-test-prefix'.
["- wrong type argument"]: New test prefix.
* test-suite/tests/numbers.test ("1+"): Use `with-test-prefix/c&e'
instead of `with-test-prefix'. Provide a name to each `pass-if'
invocation.
("1-"): Likewise.
* libguile/numbers.c (scm_iuint2str): Add an assertion on the domain of
the radix. Use the number_chars table to write the string, instead of
doing strange math. Same effect, though.
(mem2uinteger, char_decimal_value): Change logic to allow all ascii
alphabetic chars as decimals, not just a-f. Thanks to Nils Gey for the
report.
* test-suite/tests/numbers.test ("number->string"): Add some tests.
* libguile/vm-i-scheme.c (INUM_MAX, INUM_MIN): New macros.
(add1, sub1): Add/subtract without untagging the operand. This leads
to a 44% run time improvement compared to the previous
implementation.
* libguile/vm.c: Include <stdint.h>.
* test-suite/tests/numbers.test ("1+", "1-"): Add tests for
MOST-POSITIVE-FIXNUM, resp. MOST-NEGATIVE-FIXNUM, for 32-bit and
34-bit values thereof.
* benchmark-suite/benchmarks/arithmetic.bm: New file.
* benchmark-suite/Makefile.am (SCM_BENCHMARKS): Add it.
* libguile/numbers.c (scm_integer_expt): Validate the first arg as a
number.
(scm_expt): Delegate to scm_integer_expt iff x is exact. Fixes
fractions.test, which I broke recently
* test-suite/tests/numbers.test ("integer-expt"): Add test for
(integer-expt #t 0).
* libguile/tags.h: Remove rpsubrs (I chose to interpret the terse name
as "recursive predicate subrs"). Just use gsubrs with rest arguments,
or do a fold yourself.
* libguile/array-map.c (scm_i_array_equal_p): Do the comparison in
order, why not.
* libguile/chars.c:
* libguile/eq.c:
* libguile/numbers.c:
* libguile/strorder.c: Add 0,2,1 gsubr wrappers for rpsubrs like eq?, <,
etc.
* libguile/goops.c (scm_class_of)
* libguile/procprop.c (scm_i_procedure_arity)
* libguile/procs.c (scm_thunk_p)
* libguile/vm.c (apply_foreign): Remove rpsubr cases.
* test-suite/tests/numbers.test ("=", "<"): Turn a couple xfails into
passes.
Ports are given two additional properties: a character encoding and
a conversion failure strategy. These properties have getters and setters.
The new properties are used to convert any locale text to/from the
internal representation of strings.
If unspecified, ports use a default value. The default value of these
properties is held in a fluid. The default character encoding can be
modified by calling setlocale.
ISO-8859-1 is treated specially. Since it is a native encoding of
strings, it can be processed more quickly. Source code is assumed to be
ISO-8859-1 unless otherwise specified. The encoding of a source code
file can be given as 'coding: XXXXX' in a magic comment at the top of a
file.
The C functions that deal with encoding often use a null pointer
as shorthand for the native Latin-1 encoding, for efficiency's sake.
* test-suite/tests/encoding-iso88591.test: new tests
* test-suite/tests/encoding-iso88597.test: new tests
* test-suite/tests/encoding-utf8.test: new tests
* test-suite/tests/encoding-escapes.test: new tests
* test-suite/tests/numbers.test: declare 'binary' encoding
* test-suite/tests/ports.test: declare 'binary' encoding
* test-suite/tests/r6rs-ports.test: declare 'binary' encoding
* module/system/base/compile.scm (compile-file): use source-code
file's self-declared encoding when compiling files
* libguile/strports.c: store string ports in locale encoding
(scm_strport_to_locale_u8vector, scm_call_with_output_locale_u8vector)
(scm_open_input_locale_u8vector, scm_get_output_locale_u8vector):
new functions
* libguile/strings.h: new declaration for scm_i_string_contains_char
* libguile/strings.c (scm_i_string_contains_char): new function
(scm_from_stringn, scm_to_stringn): use NULL for Latin-1
(scm_from_locale_stringn, scm_to_locale_stringn): respect character
encoding of input and output ports
* libguile/read.h: declaration for scm_scan_for_encoding
* libguile/read.c:
(read_token): now takes scheme string instead of C string/length
(read_complete_token): new function
(scm_read_sexp, scm_read_number, scm_read_mixed_case_symbol)
(scm_read_number_and_radix, scm_read_quote, scm_read_semicolon_comment)
(scm_read_srfi4_vector, scm_read_bytevector, scm_read_guile_bit_vector)
(scm_read_scsh_block_comment, scm_read_commented_expression)
(scm_read_extended_symbol, scm_read_sharp_extension, scm_read_shart)
(scm_read_expression): use scm_t_wchar for char type, use read_complete_token
(scm_scan_for_encoding): new function to find a file's character encoding
(scm_file_encoding): new function to find a port's character encoding
* libguile/rdelim.c: don't unpack strings
* libguile/print.h: declaration for modified function
scm_i_charprint
* libguile/print.c: use locale when printing characters and
strings
(scm_i_charprint): input parameter is now scm_t_wchar
(scm_simple_format): don't unpack strings
* libguile/posix.h: new declaration for scm_setbinary.
* libguile/posix.c (scm_setlocale): set default and stdio port
encodings based on the locale's character encoding
(scm_setbinary): new function
* libguile/ports.h (scm_t_port): add encoding and failed
conversion handler to port type. Declarations for new or modified
functions scm_getc, scm_unget_byte, scm_ungetc,
scm_i_get_port_encoding, scm_i_set_port_encoding_x,
scm_port_encoding, scm_set_port_encoding_x,
scm_i_get_conversion_strategy, scm_i_set_conversion_strategy_x,
scm_port_conversion_strategy, scm_set_port_conversion_strategy_x.
* libguile/ports.c: assign the current ports to zero on startup so
we can see if they've been set.
(scm_current_input_port, scm_current_output_port,
scm_current_error_port): return #f if the port is not yet
initialized
(scm_new_port_table_entry): set up a new port's encoding and
illegal sequence handler based on the thread's current defaults
(scm_i_remove_port): free port encoding name when port is removed
(scm_i_mode_bits_n): now takes a scheme string instead of a c
string and length. All callers changed.
(SCM_MBCHAR_BUF_SIZE): new const
(scm_getc): new function, since the scm_getc in inline.h is now
scm_get_byte_or_eof. This pulls one codepoint from a port.
(scm_lfwrite_substr, scm_lfwrite_str): now uses port's encoding
(scm_unget_byte): new function, incorportaing the low-level functionality
of scm_ungetc
(scm_ungetc): uses scm_unget_byte
* libguile/numbers.h (scm_t_wchar): compilation order problem with
scm_t_wchar being use in functions in multiple headers. Forward
declare scm_t_wchar.
* libguile/load.c (scm_primitive_load): scan for file encoding at
top of file and use it to set the load port's encoding
* libguile/inline.h (scm_get_byte_or_eof): new function
incorporating most of the functionality of scm_getc.
* libguile/fports.c (fport_fill_input): now returns scm_t_wchar
* libguile/chars.h (scm_t_wchar): avoid compilation order problem
with declaration of scm_t_wchar
Thanks to Bill Schottstaedt for reporting this problem!
* libguile/numbers.c (mem2ureal): Don't be misled by *p_exactness
being INEXACT on entry (as is possible when reading a complex
number): use local exactness variable x which starts as EXACT.
Call mem2decimal_from_point () with &x instead of p_exactness.
* test-suite/tests/numbers.test ("string->number"): Add complex number
tests suggested by Bill.
(reported by Bill Schottstaedt)
* libguile/numbers.c (scm_gcd): When only one arg given, use scm_abs
to ensure that result is non-negative.
* test-suite/tests/numbers.test ("gcd"): New test, (gcd -2).
