diegoarjz / pagoda Goto Github PK

Exceptions are classes which are only used in a single cpp file. In that case, they should be declared and implemented in the cpp file.

See commit 402d669 for an example.

Variadic arguments in macros don't seem to be very portable and are causing travis builds to fail for gcc.

The split operation partitions geometries along one scope axis into multiple geometries.

Currently profiling and assertions are always enabled, unnecessarily slowing down the code. This should be disabled by default at least for release builds. Cmake should allow enabling this.

Implement an operation to project texture coordinates to vertices based on the scope.

Parameters:

• axis: Scope axis along which to project the coordinates.
• scale_u: How much to scale along the horizontal axis.
• scale_v: How much to scale along the vertical axis.
• offset_u: How much to translate along the horizontal axis.
• offset_v: How much to translate along the vertical axis.

The Pitchfork Layout model seems a good fit for this project.

Examples are:

• Referencing nodes by name that don't exist.
• Unknown node types
• Unknown operations
• Syntax errors

Some matrix arithmetics have been implemented but don't have unit tests associated.

Either from command line, from a file or both

The following regressions tests are failing in Ubuntu. All involve plane clipping.

RegressionTestCase.clip_geometry
RegressionTestCase.repeat_split
RegressionTestCase.split

The repeat split operation performs multiple splits of the same size along a scope axis.

This should be done as described in Pagoda's Project Layout documentation

The extract faces operation splits a geometry into its different faces.

The procedural graph scheduler is not executing the nodes in a correct order.
So far the breadth first approach has worked because there was a pattern in the way nodes were linked.

E.g.:

op1 -> out_interface1 -> in_interface1 -> op2

which balanced the nodes and they were executed correctly.

With the ParameterNode things like the following can happen, where op2 can be executed before any procedural object gets there.

par -> op2
op1 -> out_interface1 -> in_interface1 -> op2

The current graph reader implementation is made up of remnants of first experiments with selscript and doesn’t support run time expressions.

Execution of expressions is possible through selscript.

In at least one Kubuntu environment, there have been linking issues when building the unit tests:

/usr/bin/ld: ../../libgtest.a(gtest-all.cc.o): in function `testing::internal::ThreadLocal<testing::TestPartResultReporterInterface*>::~ThreadLocal()':
gtest-all.cc:(.text._ZN7testing8internal11ThreadLocalIPNS_31TestPartResultReporterInterfaceEED2Ev[_ZN7testing8internal11ThreadLocalIPNS_31TestPartResultReporterInterfaceEED5Ev]+0x24): undefined reference to `pthread_getspecific'

The parameter node should allow defining new parameters.
The node should also allow pushing the parameters into its output nodes, overwriting parameters with the same name
This way it is possible to easily define graph level parameters.

The router node allows to filter and route incoming procedural objects to different outgoing nodes.

Obj extracted from the banner regression test:

v 8.57143 0.5 0
n 0 0 1
v 8.57143 -2 0
n 0 0 1
v 10.7143 0.5 0
n 0 0 0
v 10.7143 -2 0

The way the code is structured around in math lib creates a lot of circular dependencies. This makes it harder to use.

Should reimplement the triangulation with an algorithm that supports concave polygons.