File 2075-erts-Fix-float_to_list-F-decimals-D.patch of Package erlang
From 9091e0fc630913999e3a51e860a2c4e07c6143c1 Mon Sep 17 00:00:00 2001
From: Sverker Eriksson <sverker@erlang.org>
Date: Fri, 16 Mar 2018 19:34:42 +0100
Subject: [PATCH 1/2] erts: Fix float_to_list(F, [{decimals,D}])
to better conform with io_lib:format("~.*f", [D,F])
---
erts/emulator/sys/common/erl_sys_common_misc.c | 184 +++++++++----------------
erts/emulator/test/num_bif_SUITE.erl | 25 ++++
2 files changed, 91 insertions(+), 118 deletions(-)
diff --git a/erts/emulator/sys/common/erl_sys_common_misc.c b/erts/emulator/sys/common/erl_sys_common_misc.c
index 96bdbacb9e..826307c077 100644
--- a/erts/emulator/sys/common/erl_sys_common_misc.c
+++ b/erts/emulator/sys/common/erl_sys_common_misc.c
@@ -142,7 +142,16 @@ sys_double_to_chars(double fp, char *buffer, size_t buffer_size)
return sys_double_to_chars_ext(fp, buffer, buffer_size, SYS_DEFAULT_FLOAT_DECIMALS);
}
-/* Convert float to string using fixed point notation.
+
+#if SIZEOF_LONG == 8
+# define round_int64 lround
+#elif SIZEOF_LONG_LONG == 8
+# define round_int64 llround
+#else
+# error "No 64-bit integer type?"
+#endif
+
+/* Convert float to string
* decimals must be >= 0
* if compact != 0, the trailing 0's will be truncated
*/
@@ -154,80 +163,35 @@ sys_double_to_chars_fast(double f, char *buffer, int buffer_size, int decimals,
#define FRAC_SIZE 52
#define EXP_SIZE 11
#define EXP_MASK (((Uint64)1 << EXP_SIZE) - 1)
- #define MAX_DECIMALS (sizeof(cs_sys_double_pow10) \
- / sizeof(cs_sys_double_pow10[0]))
+ #define MAX_DECIMALS (sizeof(pow10v) / sizeof(pow10v[0]))
#define FRAC_MASK (((Uint64)1 << FRAC_SIZE) - 1)
#define FRAC_MASK2 (((Uint64)1 << (FRAC_SIZE + 1)) - 1)
#define MAX_FLOAT ((Uint64)1 << (FRAC_SIZE+1))
- static const double cs_sys_double_pow10[] = {
- SYS_DOUBLE_RND_CONST / 1e0,
- SYS_DOUBLE_RND_CONST / 1e1,
- SYS_DOUBLE_RND_CONST / 1e2,
- SYS_DOUBLE_RND_CONST / 1e3,
- SYS_DOUBLE_RND_CONST / 1e4,
- SYS_DOUBLE_RND_CONST / 1e5,
- SYS_DOUBLE_RND_CONST / 1e6,
- SYS_DOUBLE_RND_CONST / 1e7,
- SYS_DOUBLE_RND_CONST / 1e8,
- SYS_DOUBLE_RND_CONST / 1e9,
- SYS_DOUBLE_RND_CONST / 1e10,
- SYS_DOUBLE_RND_CONST / 1e11,
- SYS_DOUBLE_RND_CONST / 1e12,
- SYS_DOUBLE_RND_CONST / 1e13,
- SYS_DOUBLE_RND_CONST / 1e14,
- SYS_DOUBLE_RND_CONST / 1e15,
- SYS_DOUBLE_RND_CONST / 1e16,
- SYS_DOUBLE_RND_CONST / 1e17,
- SYS_DOUBLE_RND_CONST / 1e18
+ static const double pow10v[] = {
+ 1e0, 1e1, 1e2, 1e3, 1e4, 1e5, 1e6, 1e7, 1e8, 1e9,
+ 1e10, 1e11, 1e12, 1e13, 1e14, 1e15, 1e16, 1e17, 1e18
};
- Uint64 mantissa, int_part, frac_part;
- int exp;
- int fbits;
- int max;
+ double af;
+ Uint64 int_part, frac_part;
int neg;
- double fr;
- union { Uint64 L; double F; } x;
char *p = buffer;
if (decimals < 0)
return -1;
- if (f >= 0) {
- neg = 0;
- fr = decimals < MAX_DECIMALS ? (f + cs_sys_double_pow10[decimals]) : f;
- x.F = fr;
- } else {
+ if (f < 0) {
neg = 1;
- fr = decimals < MAX_DECIMALS ? (f - cs_sys_double_pow10[decimals]) : f;
- x.F = -fr;
+ af = -f;
}
-
- exp = (x.L >> FRAC_SIZE) & EXP_MASK;
- mantissa = x.L & FRAC_MASK;
-
- if (exp == EXP_MASK) {
- if (mantissa == 0) {
- if (neg)
- *p++ = '-';
- *p++ = 'i';
- *p++ = 'n';
- *p++ = 'f';
- } else {
- *p++ = 'n';
- *p++ = 'a';
- *p++ = 'n';
- }
- *p = '\0';
- return p - buffer;
+ else {
+ neg = 0;
+ af = f;
}
- exp -= EXP_MASK >> 1;
- mantissa |= ((Uint64)1 << FRAC_SIZE);
-
/* Don't bother with optimizing too large numbers or too large precision */
- if (x.F > MAX_FLOAT || decimals >= MAX_DECIMALS) {
+ if (af > MAX_FLOAT || decimals >= MAX_DECIMALS) {
int len = erts_snprintf(buffer, buffer_size, "%.*f", decimals, f);
char* p = buffer + len;
if (len >= buffer_size)
@@ -237,77 +201,61 @@ sys_double_to_chars_fast(double f, char *buffer, int buffer_size, int decimals,
p = find_first_trailing_zero(p);
*p = '\0';
return p - buffer;
- } else if (exp >= FRAC_SIZE) {
- int_part = mantissa << (exp - FRAC_SIZE);
- frac_part = 0;
- fbits = FRAC_SIZE; /* not important as frac_part==0 */
- } else if (exp >= 0) {
- fbits = FRAC_SIZE - exp;
- int_part = mantissa >> fbits;
- frac_part = mantissa & (((Uint64)1 << fbits) -1);
- } else /* if (exp < 0) */ {
- int_part = 0;
- frac_part = mantissa;
- fbits = FRAC_SIZE - exp;
- }
-
- if (!int_part) {
- if (neg)
- *p++ = '-';
- *p++ = '0';
- } else {
- int ret, i, n;
- while (int_part != 0) {
- *p++ = (char)((int_part % 10) + '0');
- int_part /= 10;
- }
- if (neg)
- *p++ = '-';
- /* Reverse string */
- ret = p - buffer;
- for (i = 0, n = ret/2; i < n; i++) {
- int j = ret - i - 1;
- char c = buffer[i];
- buffer[i] = buffer[j];
- buffer[j] = c;
- }
}
- if (decimals > 0) {
- int i;
- *p++ = '.';
+ if (decimals) {
+ double int_f = floor(af);
+ double frac_f = round((af - int_f) * pow10v[decimals]);
- max = buffer_size - (p - buffer) - 1 /* leave room for trailing '\0' */;
+ int_part = (Uint64)int_f;
+ frac_part = (Uint64)frac_f;
- if (decimals > max)
- return -1; /* the number is not large enough to fit in the buffer */
-
- max = decimals;
+ if (frac_f >= pow10v[decimals]) {
+ /* rounding overflow carry into int_part */
+ int_part++;
+ frac_part = 0;
+ }
- for (i = 0; i < max; i++) {
- if (frac_part > (ERTS_UINT64_MAX/5)) {
- frac_part >>= 3;
- fbits -= 3;
+ do {
+ if (!frac_part) {
+ do {
+ *p++ = '0';
+ } while (--decimals);
+ break;
}
+ *p++ = (char)((frac_part % 10) + '0');
+ frac_part /= 10;
+ } while (--decimals);
- /* Multiply by 10 (5*2) to extract decimal digit as integer part */
- frac_part *= 5;
- fbits--;
+ *p++ = '.';
+ }
+ else
+ int_part = (Uint64)round_int64(af);
- if (fbits >= 64) {
- *p++ = '0';
- }
- else {
- *p++ = (char)((frac_part >> fbits) + '0');
- frac_part &= ((Uint64)1 << fbits) - 1;
- }
+ if (!int_part) {
+ *p++ = '0';
+ } else {
+ do {
+ *p++ = (char)((int_part % 10) + '0');
+ int_part /= 10;
+ }while (int_part);
+ }
+ if (neg)
+ *p++ = '-';
+
+ {/* Reverse string */
+ int i = 0;
+ int j = p - buffer - 1;
+ for ( ; i < j; i++, j--) {
+ char tmp = buffer[i];
+ buffer[i] = buffer[j];
+ buffer[j] = tmp;
}
-
- /* Delete trailing zeroes */
- if (compact)
- p = find_first_trailing_zero(p);
}
+ /* Delete trailing zeroes */
+ if (compact)
+ p = find_first_trailing_zero(p);
*p = '\0';
return p - buffer;
}
diff --git a/erts/emulator/test/num_bif_SUITE.erl b/erts/emulator/test/num_bif_SUITE.erl
index 592542405f..104bd37817 100644
--- a/erts/emulator/test/num_bif_SUITE.erl
+++ b/erts/emulator/test/num_bif_SUITE.erl
@@ -213,6 +213,20 @@ fts_rand_float_decimals(N) ->
[begin
F0 = rand_float_reasonable(),
L0 = float_to_list(F0, [{decimals, D}]),
+ case conform_with_io_lib_format(F0,D) of
+ false -> ok;
+ true ->
+ IOL = lists:flatten(io_lib:format("~.*f", [D, F0])),
+ true = case L0 =:= IOL of
+ true -> true;
+ false ->
+ io:format("F0 = ~w ~w\n", [F0, <<F0/float>>]),
+ io:format("decimals = ~w\n", [D]),
+ io:format("float_to_list = ~s\n", [L0]),
+ io:format("io_lib:format = ~s\n", [IOL]),
+ false
+ end
+ end,
L1 = case D of
0 -> L0 ++ ".0";
_ -> L0
@@ -234,6 +248,17 @@ fts_rand_float_decimals(N) ->
fts_rand_float_decimals(N-1).
+conform_with_io_lib_format(_, 0) ->
+ %% io_lib:format("~.*f") does not support zero decimals
+ false;
+conform_with_io_lib_format(_, D) when D > 10 ->
+ %% Seems float_to_list gets it slightly wrong sometimes for many decimals
+ false;
+conform_with_io_lib_format(F, D) ->
+ %% io_lib:format prints '0' for input bits beyond mantissa precision
+ %% float_to_list treats those unknown input bits as if they were zeros.
+ math:log2(abs(F) * math:pow(10,D)) < 54.
+
max_diff_decimals(F, D) ->
IntBits = floor(math:log2(abs(F))) + 1,
FracBits = (52 - IntBits),
--
2.16.3
