Bug #1324
GApplicationPars only stores reals to 6 digit precision
| Status: | Closed | Start date: | 09/25/2014 | |
|---|---|---|---|---|
| Priority: | Normal | Due date: | ||
| Assigned To: |  Mayer Michael | % Done: | 100% | |
| Category: | - | |||
| Target version: | 00-09-00 | |||
| Duration: |
Description
Ran into a problem where saving a parameter in ctselect was not working. If you give a parameter a number with more that 6 digits of precision, they get rounded to 6. You can reproduce it with the following commands:
$ python
import ctools
sel = ctools.select()
sel['tmin’].real(1234567.89)
print sel['tmin’].real()
1234570.0
sel['tmin’].real(3.546546465e9)
print sel['tmin’].real()
3546550000.0
I was hoping to set tmin and tmax to something larger (i.e. some 9- or 10-digit-precision time in seconds in veritas’s MET time), but this prevents me.
Recurrence
No recurrence.
History
#1
Updated by Kelley-Hoskins Nathan over 9 years ago
- Description updated (diff)
Example should have been formatted like this instead:
$ python>>> import ctools>>> sel = ctools.select()>>> sel['tmin'].real(1234567.89)>>> print sel['tmin'].real()1234570.0>>> sel['tmin'].real(3.546546465e9)>>> print sel['tmin'].real()3546550000.0
#2
Updated by Mayer Michael over 9 years ago
- % Done changed from 0 to 20
I dived a bit into the GApplicationPar-class. As I understand, the problem occurs in the function
std::string gammalib::str(const double &value)
in GTools.C.
The problem, however, has nothing to do with the gammalib library. It is more likely a c++ feature when casting a double into a string. The same happens if you have a double and try to print it on the screen with
std::cout<<double<<std::endl;
Luckily, gammalib::str() can take the precision as an argument. Therefore, I guess for the time cut parameters (maybe for any parameter?) we could change line 423 of GApplicationPar.C to something like
std::string value_string = gammalib::str(value,5);
or any other value. This should quickly solve this issue. In the long run, we might think about handling the time cuts with a different precision than e.g. energy or coordinate cuts.
#3
Updated by Knödlseder Jürgen over 9 years ago
I changed the code in GApplicationPar::real(double) as follows:
// Set value string at highest precision
std::string value_string = gammalib::str(value, 15);
// Strip trailing zeros
std::string::size_type start = 0;
std::string::size_type stop = value_string.find_last_not_of("0");
if (stop != std::string::npos) {
if (start <= stop) {
std::string tmp = value_string.substr(start, stop-start+1);
value_string = tmp;
}
}
The 15 is the maximum allowed precision for a double precision variable. The remaining code strips any trailing zeros from the string so that we don’t store always many zero digits for a real variable in the string.
I merged the code into the devel branch. Please let me know if this gives reasonable results on your side.
#4
Updated by Knödlseder Jürgen over 9 years ago
- Status changed from New to Feedback
- % Done changed from 20 to 80
#5
Updated by Knödlseder Jürgen over 9 years ago
- Project changed from ctools to GammaLib
#6
Updated by Knödlseder Jürgen over 9 years ago
- Assigned To set to Mayer Michael
- Target version set to 00-09-00
#7
Updated by Kelley-Hoskins Nathan over 9 years ago
- Status changed from Feedback to Resolved
- % Done changed from 80 to 100
Pulled devel and tested it, real parameters are stored to 15 digits now. Thanks.
#8
Updated by Knödlseder Jürgen over 9 years ago
- Status changed from Resolved to Closed