Float<float> and Float<double> used to share the same expression type
(Float), and the distinction was made by comparing their size. However,
due to padding, their size could be the same, leading to some issues :
- On the simulator, in Calculation, type 1_mm^3 and go to the
additional outputs. One of the results would be -0.0003081979_µm^3.
Change-Id: Ic8f9328bf462104776fbab636c34d0d152cd7e58
To prevent incorrect approximations, such as cos(1.5707963267949) = 0, we lowered the precision value. This way,
the approximation is more selective. However, when ploting functions such as e^(i.pi+x), the float approximation fails
and therefore, the function appears "undef".
As a result we created two functions Epsilon that behave differently according to the number's type. When it is a double,
we want a maximal precision -> epsilon_double = 1x10^(-15), and when it is a float, we accept more agressive approximations
-> epsilon_float = 10 x 1x10^(-7).
Change-Id: I844ac52ade665f51fe6888db38f4485c193286d9
Removed limitation in calculation of normal_distribution. It was used to
speed-up computation but yielded 0 for small values.
Previously P(X<5) with X->N(7,0.3162) gave 0 as a result
now it gives 1.26e-10
Change-Id: I3f1c997dfe2ba6424b372a0a82af6d9871443657
Changed the look of the integral symbol and the positions of the upper
and lower bounds. Added a drawing of the window to explain the meaning
of various margins.
Change-Id: I3ffc07848dc9ad68dce3c0e01853d548e31dd152
Parenthesis collapsing is no longer handled by LeftParenthesisLayout and
RightParenthesisLayout, but by ParenthesisLayout itself.
Change-Id: I6797cd40ee02881544e01e9b88f31a597c51d179
Fraction are now only collapsed if they are the first thing to be
collapsed or if a parenthesis is open.
Change-Id: Ie7f3a99498b9981dff70221efd7de6bee6d0c379
Created function Layout::LayoutFromAddress, which creates in the pool a layout
that has been stored in a buffer. Functionally identical to
Expression::ExpressionFromAddress.
Change-Id: I204e6b7d42f259ef2732563aa2a168a644060867
Derivation now computes as expected on logarithms, as long as their base is not a function of the derivation variable.
Change-Id: Ia56da1c1151c0ddf3887be84ddb4bd02664c5188
Added a method unaryFunctionDifferential to ExpressionNode and Expression, to be implemented by subclasses representing unary functions.
Added a function derivateUnaryFunction to Derivative, to factor (f°g)' -> g' * f'°g.
Change-Id: Id1780f1082ccd001f1282fe4ddfff2b7055d3a27
Derivation now propagates to powers as expected, whether the base, the exponent, or both are functions of the variable. This also makes division derive as intended.
Change-Id: I51cbd5f7ec9f6aaa1df068625bbda1437941fa08
Derivation propagates as expected on multiplications (but not power). Some tests involving diff had to be updated to reflect that behaviour.
Change-Id: Ifa32031bc37a156c18d296757bcdd6ccdb0ea43e
Number's subclasses, with the exception of Undefined and Unreal, derive as zero. Symbols derive as expected. Derivation propagates as expected on additions.
Change-Id: Icedbb1dfac8099347a19c14bf2537011b2e8b199
This method is to be implemented by derivable expression subclasses, for use in Derivative::shallowReduce. It performs the calculation for the derivative, and returns whether calculations happened.
Change-Id: I13cdb131e2044578392f5178a9f389314c1c4c8a
the object are ordered in a table. Otherwise, the compiler is free to
order them as it wants.
This fixes the test on the device: 10_m^2→_mm×_km simplifies to 10×_km×_mm
