Evaluating a function containing a sequence, an integral or a
derivative, is time consuming. Computing a range requires a plethora of
evaluations, as such ranges for these functions cannot be evaluated in a
timely fashion.
Change-Id: I088a0e896dbc26e6563291cafdfe9ceba36dd5d0
When building an orthonormal range for the automatic zoom, we provide a
ratio for the screen without the margins. This way, adding the margins
will make the graph orthonormal, and the banner cannot cover the
function.
Change-Id: If3a3f799d4e7e3e81ab77c6b418d70b734a6fbca
When a function that had previously been cached is deactivated, its
cache can be used by another function. When the function is reactivated,
if it tries to reuse its previous cache, it will be filled with values
from another function, and will not be cleared. By detaching the cache
of a function that becomes inactive, we ensure that the next time this
function is cached, the chache will be cleared.
e.g. Define to functions f(x)=x, g(x)=-x, draw them both.
Deactivate f, the redraw the graph
Reactivate f, the draw again
The ability to perform a range search without changing the X range was
only there to accomodate the Yauto feature.
Change-Id: I8c2b61b447fbd3dc1f4e303dff06d1a8d6e7a4f2
The new zoom implemented for ContinuousFunction is now factorized inside
Function to benefit the Sequence class. The same things is done to code
added to Graph::GraphController, which is moved into
FunctionGraphController.
This removes the reimplementation of several methods, most notably
computeYRange, as the implementation for function is general enough to
work on sequences.
Change-Id: I9b8211354064f46c3fa3dde3191dcb39d627a1d2
Initial zoom for displaying curves is computed according to the
following rules :
- For polar and parametric curves, the algorithm has not changed.
Vertical and horizontal ranges are chosen in order to display the
full curve with orthonormal graduations.
- For cartesian curves, the horizontal range is chosen in order to
display all points of interest (roots, extrema, asymptotes...).
The vertical range is fitted to be able to display those points, but
also cuts out points outside of the function's order of magnitude.
Change-Id: Idf8233fc2e6586b85d34c4da152c83e75513d85c
Instead of being memoized all at once before display, functions values
are now stored at evaluation time. Fixed some quirks with caching
preparation.
Change-Id: I5d212c271c8c41a6dc9074a15c720f0bccf8ac40
The caches used for function values memoization are now stored in
ContinuousFunctionStore : there are now only a fixed number, instead of
one per function. This effectively enables caching only for the first
few functions on screen, while reducing the memory usage.
Change-Id: I2ade091717f73a14a756fe527c773db8e8627be7
When evaluating a ContinuousFunction for a float value, the function
will first try to ask its cache (if it has been filled beforehand).
Change-Id: I519d2d3dcf344ba63e30a0f011db2306c7141315
ContinuousFunction now has an attribute of type ContinuousFunctionCache,
implementing methods to store and retrieve 320 float values, in order to
speed up function display in Graph.
Change-Id: I6f7ccdf3ae3c6dd8b08b93d786c8d0be7aa4dee8
It appeared that without the packed keyword, the compiler did not handle
access to unaligned record data members, which leads to crashes on the
device.
Change-Id: I401f075e7f62458a4733baa8d81983d4be34b730
