MCGR: ColorSteps handling moved to tooling

Due to the fact that handling of a vector GradientSteps
will get used more often with ongoing changes I moved
some handling to tooling, see sortAndCorrectColorSteps.
That method sorts and corrects a vector of ColorSteps
so that only valid entreis remain in a sorted order,
for details please refer to the docu at function
definition. I took the occasion to rearrange that to
work on the single provided vector which is better for
speed & ressources.

Also changed the ColorStep constructor to not
automatically correct constructed entries: While that
is formally correct, it moves an invalid entry to 0.0
or 1.0, thus creating additional wrong Start/EndColor
entries. Those may then 'overlay' the correct entry
when corrections are applied to the vector of entries
which leads to getting the wanted Start/EndColor to be
factically deleted, what is an error.

Also rearranged FillGradientAttribute to now work
initially with ColorSteps, no longer requires the
Start/EndColor, also adapted all usages. This securely
allows start/end and in-between single-color sections
in gradients.

Also needed to re-formulate GradientRect &
GradientElliptical ::appendTransformationsAndColors
method for correct 2D mapping(s) - always incrementing
from the start to end conflicts with the adaption of
the start value for the inner loop mainly because
increment is at start of inner loop (pre-increment).

Corrected errors from clang plugin, but also some
wrong initialiations for basegfx::ColorSteps.

Change-Id: I292592ed9abcfa591f68a680479f4fcdda46cbeb
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/148196
Tested-by: Jenkins
Reviewed-by: Armin Le Grand <Armin.Le.Grand@me.com>

1 files changed, 30 insertions, 1 deletions

diff --git a/include/basegfx/utils/gradienttools.hxx b/include/basegfx/utils/gradienttools.hxx
index 793d07a75577..328a86b8335c 100644
--- a/include/basegfx/utils/gradienttools.hxx
+++ b/include/basegfx/utils/gradienttools.hxx
@@ -69,9 +69,18 @@ namespace basegfx
         // e.g. for usage in std::vector::insert
         // ensure [0.0 .. 1.0] range for mfOffset
         ColorStep(double fOffset = 0.0, const BColor& rColor = BColor())
-            : mfOffset(std::max(0.0, std::min(fOffset, 1.0)))
+            : mfOffset(fOffset)
             , maColor(rColor)
         {
+            // NOTE: I originally *corrected* mfOffset here by using
+            //   mfOffset(std::max(0.0, std::min(fOffset, 1.0)))
+            // While that is formally correct, it moves an invalid
+            // entry to 0.0 or 1.0, thus creating additional wrong
+            // Start/EndColor entries. That may then 'overlay' the
+            // cortrect entry when corrections are applied to the
+            // vector of entries (see sortAndCorrectColorSteps)
+            // which leads to getting the wanted Start/EndColor
+            // to be factically deleted, what is an error.
         }
 
         double getOffset() const { return mfOffset; }
@@ -184,6 +193,26 @@ namespace basegfx
 
     namespace utils
     {
+        /* Helper to sort and correct ColorSteps. This will
+           sort and then correct the given ColorSteps. The
+           corrected version will
+           - be sorted
+           - have no double values
+           - have no values with offset < 0.0
+           - have no values with offset > 1.0
+           thus be ready to be used in multi-color gradients.
+
+           NOTE: The returned version may be empty (!) if no
+                 valid entries were contained
+           NOTE: It does not necessarily contain values for
+                 offset == 0.0 and 1.0 if there were none
+                 given (so no Start/EndColor)
+           NOTE: If it contains only one entry that entry is
+                 set to StartColor and the Color is preserved.
+                 This is also done when all Colors are the same
+        */
+        BASEGFX_DLLPUBLIC void sortAndCorrectColorSteps(ColorSteps& rColorSteps);
+
         /* Helper to grep the correct ColorStep out of
            ColorSteps and interpolate as needed for given
            relative value in fScaler in the range of [0.0 .. 1.0].