File 0346-Fix-negative-zero-to-string.patch of Package erlang
From e78af03b493413214816a58dbf4845a01120a4ea Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Jos=C3=A9=20Valim?= <jose.valim@dashbit.co>
Date: Wed, 2 Dec 2020 16:04:59 +0100
Subject: [PATCH] Fix negative zero to string
Both float_to_binary and float_to_list will correctly
print negative zeros when using the scientific mode but
not when using the decimal representation. This patch
makes it consistent to always show negative floats.
io:format/2 has also been fixed to consider negative zeros.
This is important because some operations, such as math:atan2,
will return drastically different results if negative zeros
are given compared to positive zeros. Negative zeros also
are encoded differently in bitstrings.
---
.../emulator/sys/common/erl_sys_common_misc.c | 2 +-
erts/emulator/test/num_bif_SUITE.erl | 11 ++++++
lib/stdlib/src/io_lib_format.erl | 38 +++++++++++++------
lib/stdlib/test/io_SUITE.erl | 25 +++++++++---
4 files changed, 58 insertions(+), 18 deletions(-)
diff --git a/erts/emulator/sys/common/erl_sys_common_misc.c b/erts/emulator/sys/common/erl_sys_common_misc.c
index fd7c3b2cda..c840971e6d 100644
--- a/erts/emulator/sys/common/erl_sys_common_misc.c
+++ b/erts/emulator/sys/common/erl_sys_common_misc.c
@@ -182,7 +182,7 @@ sys_double_to_chars_fast(double f, char *buffer, int buffer_size, int decimals,
if (decimals < 0)
return -1;
- if (f < 0) {
+ if (signbit(f)) {
neg = 1;
af = -f;
}
diff --git a/erts/emulator/test/num_bif_SUITE.erl b/erts/emulator/test/num_bif_SUITE.erl
index 6b834705cf..f055a4d7f8 100644
--- a/erts/emulator/test/num_bif_SUITE.erl
+++ b/erts/emulator/test/num_bif_SUITE.erl
@@ -187,6 +187,17 @@ t_float_to_string(Config) when is_list(Config) ->
test_fts("1.2300000000e+20",1.23e20, [{scientific, 10}, compact]),
test_fts("1.23000000000000000000e+20",1.23e20, []),
+ %% Negative zero
+ <<NegZero/float>> = <<16#8000000000000000:64>>,
+ "-0.0" = float_to_list(NegZero, [{decimals, 1}, compact]),
+ "-0.0" = float_to_list(NegZero, [{decimals, 1}]),
+ "-0.0e+00" = float_to_list(NegZero, [{scientific, 1}]),
+ "-0.0e+00" = float_to_list(NegZero, [{scientific, 1}, compact]),
+ <<"-0.0">> = float_to_binary(NegZero, [{decimals, 1}, compact]),
+ <<"-0.0">> = float_to_binary(NegZero, [{decimals, 1}]),
+ <<"-0.0e+00">> = float_to_binary(NegZero, [{scientific, 1}]),
+ <<"-0.0e+00">> = float_to_binary(NegZero, [{scientific, 1}, compact]),
+
fts_rand_float_decimals(1000),
ok.
diff --git a/lib/stdlib/src/io_lib_format.erl b/lib/stdlib/src/io_lib_format.erl
index 435b31750e..bf6869a492 100644
--- a/lib/stdlib/src/io_lib_format.erl
+++ b/lib/stdlib/src/io_lib_format.erl
@@ -449,9 +449,10 @@ fwrite_e(Fl, F, Adj, none, Pad) ->
fwrite_e(Fl, F, Adj, P, Pad) when P >= 2 ->
term(float_e(Fl, float_data(Fl), P), F, Adj, F, Pad).
-float_e(Fl, Fd, P) when Fl < 0.0 -> %Negative numbers
- [$-|float_e(-Fl, Fd, P)];
-float_e(_Fl, {Ds,E}, P) ->
+float_e(Fl, Fd, P) ->
+ signbit(Fl) ++ abs_float_e(abs(Fl), Fd, P).
+
+abs_float_e(_Fl, {Ds,E}, P) ->
case float_man(Ds, 1, P-1) of
{[$0|Fs],true} -> [[$1|Fs]|float_exp(E)];
{Fs,false} -> [Fs|float_exp(E-1)]
@@ -503,16 +504,27 @@ fwrite_f(Fl, F, Adj, none, Pad) ->
fwrite_f(Fl, F, Adj, P, Pad) when P >= 1 ->
term(float_f(Fl, float_data(Fl), P), F, Adj, F, Pad).
-float_f(Fl, Fd, P) when Fl < 0.0 ->
- [$-|float_f(-Fl, Fd, P)];
-float_f(Fl, {Ds,E}, P) when E =< 0 ->
- float_f(Fl, {lists:duplicate(-E+1, $0)++Ds,1}, P); %Prepend enough 0's
-float_f(_Fl, {Ds,E}, P) ->
+float_f(Fl, Fd, P) ->
+ signbit(Fl) ++ abs_float_f(abs(Fl), Fd, P).
+
+abs_float_f(Fl, {Ds,E}, P) when E =< 0 ->
+ abs_float_f(Fl, {lists:duplicate(-E+1, $0)++Ds,1}, P); %Prepend enough 0's
+abs_float_f(_Fl, {Ds,E}, P) ->
case float_man(Ds, E, P) of
{Fs,true} -> "1" ++ Fs; %Handle carry
{Fs,false} -> Fs
end.
+%% signbit(Float) -> [$-] | []
+
+signbit(Fl) when Fl < 0.0 -> [$-];
+signbit(Fl) when Fl > 0.0 -> [];
+signbit(Fl) ->
+ case <<Fl/float>> of
+ <<1:1,_:63>> -> [$-];
+ _ -> []
+ end.
+
%% float_data([FloatChar]) -> {[Digit],Exponent}
float_data(Fl) ->
@@ -545,13 +557,15 @@ float_data([_|Cs], Ds) ->
-spec fwrite_g(float()) -> string().
-fwrite_g(0.0) ->
+fwrite_g(Fl) ->
+ signbit(Fl) ++ abs_fwrite_g(abs(Fl)).
+
+abs_fwrite_g(0.0) ->
"0.0";
-fwrite_g(Float) when is_float(Float) ->
+abs_fwrite_g(Float) when is_float(Float) ->
{Frac, Exp} = mantissa_exponent(Float),
{Place, Digits} = fwrite_g_1(Float, Exp, Frac),
- R = insert_decimal(Place, [$0 + D || D <- Digits], Float),
- [$- || true <- [Float < 0.0]] ++ R.
+ insert_decimal(Place, [$0 + D || D <- Digits], Float).
-define(BIG_POW, (1 bsl 52)).
-define(MIN_EXP, (-1074)).
diff --git a/lib/stdlib/test/io_SUITE.erl b/lib/stdlib/test/io_SUITE.erl
index 5b8e42db1c..bb59a5e53f 100644
--- a/lib/stdlib/test/io_SUITE.erl
+++ b/lib/stdlib/test/io_SUITE.erl
@@ -29,7 +29,7 @@
printable_range/1, bad_printable_range/1,
io_lib_print_binary_depth_one/1, otp_10302/1, otp_10755/1,
otp_10836/1, io_lib_width_too_small/1,
- io_with_huge_message_queue/1, format_string/1,
+ io_with_huge_message_queue/1, format_string/1, format_neg_zero/1,
maps/1, coverage/1, otp_14178_unicode_atoms/1, otp_14175/1,
otp_14285/1, limit_term/1, otp_14983/1, otp_15103/1, otp_15076/1,
otp_15159/1, otp_15639/1, otp_15705/1, otp_15847/1, otp_15875/1,
@@ -60,7 +60,7 @@ all() ->
manpage, otp_6708, otp_7084, otp_7421,
io_lib_collect_line_3_wb, cr_whitespace_in_string,
io_fread_newlines, otp_8989, io_lib_fread_literal,
- printable_range, bad_printable_range,
+ printable_range, bad_printable_range, format_neg_zero,
io_lib_print_binary_depth_one, otp_10302, otp_10755, otp_10836,
io_lib_width_too_small, io_with_huge_message_queue,
format_string, maps, coverage, otp_14178_unicode_atoms, otp_14175,
@@ -88,6 +88,14 @@ error_1(Config) when is_list(Config) ->
{'EXIT', _} = (catch io:format(F1, "~p", ["hej"])),
ok.
+format_neg_zero(Config) when is_list(Config) ->
+ <<NegZero/float>> = <<16#8000000000000000:64>>,
+ "-0.000000" = io_lib:format("~f", [NegZero]),
+ "-0.00000e+0" = io_lib:format("~g", [NegZero]),
+ "-0.00000e+0" = io_lib:format("~e", [NegZero]),
+ "-0.0" = io_lib_format:fwrite_g(NegZero),
+ ok.
+
float_g(Config) when is_list(Config) ->
["5.00000e-2",
"0.500000",
@@ -1094,13 +1102,20 @@ g_t(V) when is_float(V) ->
%% Note: in a few cases the least significant digit has been
%% incremented by one, namely when the correctly rounded string
%% converts to another floating point number.
-g_t(0.0, "0.0") ->
- ok;
-g_t(V, Sv) ->
+g_t(V, Sv) when V > 0.0; V < 0.0 ->
try
g_t_1(V, Sv)
catch throw:Reason ->
throw({Reason, V, Sv})
+ end;
+g_t(Zero, Format) ->
+ case <<Zero/float>> of
+ <<1:1,_:63>> ->
+ "-0.0" = Format,
+ ok;
+ <<0:1,_:63>> ->
+ "0.0" = Format,
+ ok
end.
g_t_1(V, Sv) ->
--
2.26.2
