UpsilonNumworks/Upsilon
1
0
Fork 0
mirror of https://github.com/UpsilonNumworks/Upsilon.git synced 2026-01-19 00:37:25 +01:00
Code Issues Packages Projects Releases Wiki Activity

[poincare/absolute_value] ShallowReduce done in double, not float

Browse Source 
This fixes abs(-2.3*10^-39) that returned a negative value
This commit is contained in:
Léa Saviot
2020-03-04 10:17:00 +01:00
committed by RubenNumworks
parent 5ca6b7dbf8
commit cf84a30768
1 changed files with 15 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
poincare/src/absolute_value.cpp
View File

@@ -50,21 +50,21 @@ Expression AbsoluteValue::shallowReduce(ExpressionNode::ReductionContext reducti
}
// |x| = ±x if x is real
if (c.isReal(reductionContext.context())) {
float app = c.node()->approximate(float(), reductionContext.context(), reductionContext.complexFormat(), reductionContext.angleUnit()).toScalar();
if (!std::isnan(app) &&
((c.isNumber() && app >= 0) || app >= Expression::Epsilon<float>())) {
/* abs(a) = a with a >= 0
* To check that a > 0, if a is a number we can use float comparison;
* in other cases, we are more conservative and rather check that
* a > epsilon ~ 1E-7 to avoid potential error due to float precision. */
replaceWithInPlace(c);
return c;
} else if (!std::isnan(app) &&
((c.isNumber() && app < 0.0f) || app <= -Expression::Epsilon<float>())) {
// abs(a) = -a with a < 0 (same comment as above to check that a < 0)
Multiplication m = Multiplication::Builder(Rational::Builder(-1), c);
replaceWithInPlace(m);
return m.shallowReduce(reductionContext);
double app = c.node()->approximate(double(), reductionContext.context(), reductionContext.complexFormat(), reductionContext.angleUnit()).toScalar();
if (!std::isnan(app)) {
if ((c.isNumber() && app >= 0) || app >= Expression::Epsilon<double>()) {
/* abs(a) = a with a >= 0
* To check that a > 0, if a is a number we can use float comparison;
* in other cases, we are more conservative and rather check that
* a > epsilon ~ 1E-7 to avoid potential error due to float precision. */
replaceWithInPlace(c);
return c;
} else if ((c.isNumber() && app < 0.0f) || app <= -Expression::Epsilon<double>()) {
// abs(a) = -a with a < 0 (same comment as above to check that a < 0)
Multiplication m = Multiplication::Builder(Rational::Builder(-1), c);
replaceWithInPlace(m);
return m.shallowReduce(reductionContext);
}
}
}
// |a+ib| = sqrt(a^2+b^2)