Do not use scientific notations when printing doubles in logs.

Closes BIT-1558.
This commit is contained in:
Robin Sommer 2016-05-23 14:42:13 -07:00
parent 3581ead0d9
commit d86bf15dbf
7 changed files with 197 additions and 5 deletions

View file

@ -4,6 +4,7 @@
#include <stdlib.h>
#include <errno.h>
#include <math.h>
#include "Desc.h"
#include "File.h"
@ -138,17 +139,22 @@ void ODesc::Add(uint64 u)
}
}
void ODesc::Add(double d)
void ODesc::Add(double d, bool no_exp)
{
if ( IsBinary() )
AddBytes(&d, sizeof(d));
else
{
char tmp[256];
if ( no_exp )
modp_dtoa3(d, tmp, sizeof(tmp), IsReadable() ? 6 : 8);
else
modp_dtoa2(d, tmp, IsReadable() ? 6 : 8);
Add(tmp);
if ( d == double(int(d)) )
if ( nearbyint(d) == d && isfinite(d) && ! strchr(tmp, 'e') )
// disambiguate from integer
Add(".0");
}

View file

@ -81,7 +81,7 @@ public:
void Add(uint32 u);
void Add(int64 i);
void Add(uint64 u);
void Add(double d);
void Add(double d, bool no_exp=false);
void Add(const IPAddr& addr);
void Add(const IPPrefix& prefix);

View file

@ -287,5 +287,136 @@ void modp_dtoa2(double value, char* str, int prec)
strreverse(str, wstr-1);
}
// This is near identical to modp_dtoa2 above, excep that it never uses
// exponential notation and requires a buffer length.
void modp_dtoa3(double value, char* str, int n, int prec)
{
/* Hacky test for NaN
* under -fast-math this won't work, but then you also won't
* have correct nan values anyways. The alternative is
* to link with libmath (bad) or hack IEEE double bits (bad)
*/
if (! (value == value)) {
str[0] = 'n'; str[1] = 'a'; str[2] = 'n'; str[3] = '\0';
return;
}
/* if input is larger than thres_max, revert to exponential */
const double thres_max = (double)(0x7FFFFFFF);
int count;
double diff = 0.0;
char* wstr = str;
if (prec < 0) {
prec = 0;
} else if (prec > 9) {
/* precision of >= 10 can lead to overflow errors */
prec = 9;
}
/* we'll work in positive values and deal with the
negative sign issue later */
int neg = 0;
if (value < 0) {
neg = 1;
value = -value;
}
int whole = (int) value;
double tmp = (value - whole) * _pow10[prec];
uint32_t frac = (uint32_t)(tmp);
diff = tmp - frac;
if (diff > 0.5) {
++frac;
/* handle rollover, e.g. case 0.99 with prec 1 is 1.0 */
if (frac >= _pow10[prec]) {
frac = 0;
++whole;
}
} else if (diff == 0.5 && ((frac == 0) || (frac & 1))) {
/* if halfway, round up if odd, OR
if last digit is 0. That last part is strange */
++frac;
}
/* for very large numbers switch back to native sprintf for exponentials.
anyone want to write code to replace this? */
/*
normal printf behavior is to print EVERY whole number digit
which can be 100s of characters overflowing your buffers == bad
*/
if (value > thres_max) {
/* ---- Modified part, compared to modp_dtoa3. */
int i = snprintf(str, n, "%.*f", prec, neg ? -value : value);
if ( i < 0 || i >= n ) {
// Error or truncated output.
snprintf(str, n, "NAN");
return;
}
/* Remove trailing zeros. */
char* p;
for ( p = str + i - 1; p >= str && *p == '0'; --p );
if ( p >= str && *p == '.' )
--p;
*++p = '\0';
return;
/* ---- End of modified part.. */
}
if (prec == 0) {
diff = value - whole;
if (diff > 0.5) {
/* greater than 0.5, round up, e.g. 1.6 -> 2 */
++whole;
} else if (diff == 0.5 && (whole & 1)) {
/* exactly 0.5 and ODD, then round up */
/* 1.5 -> 2, but 2.5 -> 2 */
++whole;
}
//vvvvvvvvvvvvvvvvvvv Diff from modp_dto2
} else if (frac) {
count = prec;
// now do fractional part, as an unsigned number
// we know it is not 0 but we can have leading zeros, these
// should be removed
while (!(frac % 10)) {
--count;
frac /= 10;
}
//^^^^^^^^^^^^^^^^^^^ Diff from modp_dto2
// now do fractional part, as an unsigned number
do {
--count;
*wstr++ = (char)(48 + (frac % 10));
} while (frac /= 10);
// add extra 0s
while (count-- > 0) *wstr++ = '0';
// add decimal
*wstr++ = '.';
}
// do whole part
// Take care of sign
// Conversion. Number is reversed.
do *wstr++ = (char)(48 + (whole % 10)); while (whole /= 10);
if (neg) {
*wstr++ = '-';
}
*wstr='\0';
strreverse(str, wstr-1);
}

View file

@ -97,6 +97,15 @@ void modp_dtoa(double value, char* buf, int precision);
*/
void modp_dtoa2(double value, char* buf, int precision);
/** \brief convert a floating point number to char buffer with a
* variable-precision format, no trailing zeros, and no
* scientific notation.
*
* Other than avoiding scientific notation, this is the same as mop_dtoa2. It does however
* require the max buffer length. The buffer will always be null-terminated.
*/
void modp_dtoa3(double value, char* buf, int n, int precision);
END_C
#endif

View file

@ -91,7 +91,7 @@ bool Ascii::Describe(ODesc* desc, threading::Value* val, const string& name) con
// Rendering via Add() truncates trailing 0s after the
// decimal point. The difference with TIME/INTERVAL is mainly
// to keep the log format consistent.
desc->Add(val->val.double_val);
desc->Add(val->val.double_val, true);
break;
case TYPE_INTERVAL:

View file

@ -0,0 +1,17 @@
#separator \x09
#set_separator ,
#empty_field (empty)
#unset_field -
#path test
#open 2016-05-23-22-44-54
#fields d
#types double
2153226000.0
2153226000.1
2153226000.123457
1.0
1.1
1.123457
1.1234
3140000000000000.0
#close 2016-05-23-22-44-54

View file

@ -0,0 +1,29 @@
#
# @TEST-EXEC: bro -b %INPUT
# @TEST-EXEC: btest-diff test.log
#
# Make sure we do not write out scientific notation for doubles.
module Test;
export {
redef enum Log::ID += { LOG };
type Info: record {
d: double &log;
};
}
event bro_init()
{
Log::create_stream(Test::LOG, [$columns=Info]);
Log::write(Test::LOG, [$d=2153226000.0]);
Log::write(Test::LOG, [$d=2153226000.1]);
Log::write(Test::LOG, [$d=2153226000.123456789]);
Log::write(Test::LOG, [$d=1.0]);
Log::write(Test::LOG, [$d=1.1]);
Log::write(Test::LOG, [$d=1.123456789]);
Log::write(Test::LOG, [$d=1.1234]);
Log::write(Test::LOG, [$d=3.14e15]);
}