GeomVertexData

class GeomVertexData

Bases: CopyOnWriteObject, GeomEnums

This defines the actual numeric vertex data stored in a Geom, in the structure defined by a particular GeomVertexFormat object.

The data consists of one or more arrays, each of which in turn consists of a series of rows, one per vertex. All arrays should have the same number of rows; each vertex is defined by the column data from a particular row across all arrays.

Often, there will be only one array per Geom, and the various columns defined in the GeomVertexFormat will be interleaved within that array. However, it is also possible to have multiple different arrays, with a certain subset of the total columns defined in each array.

However the data is distributed, the effect is of a single table of vertices, where each vertex is represented by one row of the table.

In general, application code should not attempt to directly manipulate the vertex data through this structure; instead, use the GeomVertexReader, GeomVertexWriter, and GeomVertexRewriter objects to read and write vertex data at a high level.

Inheritance diagram

Inheritance diagram of GeomVertexData

explicit GeomVertexData(std::string const &name, GeomVertexFormat const *format, GeomEnums::UsageHint usage_hint)
GeomVertexData(GeomVertexData const &copy)
explicit GeomVertexData(GeomVertexData const &copy, GeomVertexFormat const *format)

Constructs an invalid object. This is only used when reading from the bam file.

This constructor copies all of the basic properties of the source VertexData, like usage_hint and animation tables, but does not copy the actual data, and it allows you to specify a different format.

ConstPointerTo<GeomVertexData> animate_vertices(bool force, Thread *current_thread) const

Returns a GeomVertexData that represents the results of computing the vertex animation on the CPU for this GeomVertexData.

If there is no CPU-defined vertex animation on this object, this just returns the original object.

If there is vertex animation, but the VertexTransform values have not changed since last time, this may return the same pointer it returned previously. Even if the VertexTransform values have changed, it may still return the same pointer, but with its contents modified (this is preferred, since it allows the graphics backend to update vertex buffers optimally).

If force is false, this method may return immediately with stale data, if the vertex data is not completely resident. If force is true, this method will never return stale data, but may block until the data is available.

void clear_animated_vertices(void)

Removes the cache of animated vertices computed by a previous call to animate_vertices() within the same frame. This will force the next call to animate_vertices() to recompute these values from scratch. Normally it is not necessary to call this.

void clear_cache(void)

Removes all of the previously-cached results of convert_to().

This blows away the entire cache, upstream and downstream the pipeline. Use clear_cache_stage() instead if you only want to blow away the cache at the current stage and upstream.

void clear_cache_stage(void)

Removes all of the previously-cached results of convert_to(), at the current pipeline stage and upstream. Does not affect the downstream cache.

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

void clear_rows(void)

Removes all of the rows from the arrays; functionally equivalent to set_num_rows(0) (but faster).

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

void clear_slider_table(void)

Sets the SliderTable pointer to NULL, removing the table from the vertex data. This disables morph (blend shape) animation.

void clear_transform_blend_table(void)

Sets the TransformBlendTable pointer to NULL, removing the table from the vertex data. This disables CPU-driven vertex animation.

void clear_transform_table(void)

Sets the TransformTable pointer to NULL, removing the table from the vertex data. This disables hardware-driven vertex animation.

int compare_to(GeomVertexData const &other) const

Returns 0 if the two objects are equivalent, even if they are not the same pointer.

ConstPointerTo<GeomVertexData> convert_to(GeomVertexFormat const *new_format) const

Returns a new GeomVertexData that represents the same contents as this one, with all data types matched up name-by-name to the indicated new format.

void copy_from(GeomVertexData const *source, bool keep_data_objects, Thread *current_thread = Thread::get_current_thread())

Copies all the data from the other array into the corresponding data types in this array, by matching data types name-by-name.

keep_data_objects specifies what to do when one or more of the arrays can be copied without the need to apply any conversion operation. If it is true, the original GeomVertexArrayData objects in this object are retained, and their data arrays are copied byte-by-byte from the source; if it is false, then the GeomVertexArrayData objects are copied pointerwise from the source.

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

void copy_row_from(int dest_row, GeomVertexData const *source, int source_row, Thread *current_thread)

Copies a single row of the data from the other array into the indicated row of this array. In this case, the source format must exactly match the destination format.

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

void describe_vertex(std::ostream &out, int row) const

Writes a verbose, human-friendly description of the indicated vertex number.

ConstPointerTo<GeomVertexArrayData> get_array(std::size_t i) const

Returns a const pointer to the vertex data for the indicated array, for application code to directly examine (but not modify) the underlying vertex data.

ConstPointerTo<GeomVertexArrayDataHandle> get_array_handle(std::size_t i) const

Equivalent to get_array(i).get_handle().

static TypeHandle get_class_type(void)
GeomVertexFormat const *get_format(void) const

Returns a pointer to the GeomVertexFormat structure that defines this data.

UpdateSeq get_modified(Thread *current_thread = Thread::get_current_thread()) const

Returns a sequence number which is guaranteed to change at least every time the vertex data is modified.

std::string const &get_name(void) const

Returns the name passed to the constructor, if any. This name is reported on the PStats graph for vertex computations.

std::size_t get_num_arrays(void) const

Returns the number of individual arrays stored within the data. This must match get_format()->get_num_arrays().

int get_num_bytes(void) const

Returns the total number of bytes consumed by the different arrays of the vertex data.

int get_num_rows(void) const

Returns the number of rows stored within all the arrays. All arrays store data for the same n rows.

SliderTable const *get_slider_table(void) const

Returns a const pointer to the SliderTable assigned to this data. Vertices within the vertex data will look up their morph offsets, if any, within this table.

This will return NULL if the vertex data does not have a SliderTable assigned.

ConstPointerTo<TransformBlendTable> get_transform_blend_table(void) const

Returns a const pointer to the TransformBlendTable assigned to this data. Vertices within the table will index into this table to indicate their dynamic skinning information; this table is used when the vertex animation is to be performed by the CPU (but also see get_transform_table()).

This will return NULL if the vertex data does not have a TransformBlendTable assigned (which implies the vertices will not be animated by the CPU).

TransformTable const *get_transform_table(void) const

Returns a const pointer to the TransformTable assigned to this data. Vertices within the table will index into this table to indicate their dynamic skinning information; this table is used when the vertex animation is to be performed by the graphics hardware (but also see get_transform_blend_table()).

This will return NULL if the vertex data does not have a TransformTable assigned (which implies the vertices will not be animated by the graphics hardware).

GeomEnums::UsageHint get_usage_hint(void) const

Returns the usage hint that was passed to the constructor, and which will be passed to each array data object created initially, and arrays created as the result of a convert_to() operation. See geomEnums.h.

However, each individual array may be replaced with a different array object with an independent usage hint specified, so there is no guarantee that the individual arrays all have the same usage_hint.

bool has_column(InternalName const *name) const

Returns true if the data has the named column, false otherwise. This is really just a shortcut for asking the same thing from the format.

void insert_array(std::size_t i, GeomVertexArrayData const *array)

Inserts the indicated vertex data array into the list of arrays, which also modifies the format. You should be careful that the new array has the same number of rows as the vertex data.

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

PointerTo<GeomVertexArrayData> modify_array(std::size_t i)

Returns a modifiable pointer to the indicated vertex array, so that application code may directly manipulate the data. You should avoid changing the length of this array, since all of the arrays should be kept in sync–use set_num_rows() instead.

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

PointerTo<GeomVertexArrayDataHandle> modify_array_handle(std::size_t i)

Equivalent to modify_array(i).modify_handle().

PointerTo<TransformBlendTable> modify_transform_blend_table(void)

Returns a modifiable pointer to the current TransformBlendTable on this vertex data, if any, or NULL if there is not a TransformBlendTable. See get_transform_blend_table().

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

void output(std::ostream &out) const
void remove_array(std::size_t i)

Removes the array with the given index from the GeomVertexData.

PointerTo<GeomVertexData> replace_column(InternalName *name, int num_components, GeomEnums::NumericType numeric_type, GeomEnums::Contents contents) const

Returns a new GeomVertexData object, suitable for modification, with the indicated data type replaced with a new table filled with undefined values. The new table will be added as a new array; if the old table was interleaved with a previous array, the previous array will not be repacked.

If num_components is 0, the indicated name is simply removed from the type, without replacing it with anything else.

bool request_resident(void) const

Returns true if the vertex data is currently resident in memory. If this returns false, the vertex data will be brought back into memory shortly; try again later.

bool reserve_num_rows(int n)

This ensures that enough memory space for n rows is allocated, so that you may increase the number of rows to n without causing a new memory allocation. This is a performance optimization only; it is especially useful when you know ahead of time that you will be adding n rows to the data.

If you know exactly how many rows you will be needing, it is significantly faster to use set_num_rows() or unclean_set_num_rows() instead.

ConstPointerTo<GeomVertexData> reverse_normals(void) const

Returns a new GeomVertexData object with the normal data modified in-place, so that each lighting normal is now facing in the opposite direction.

If the vertex data does not include a normal column, this returns the original GeomVertexData object, unchanged.

ConstPointerTo<GeomVertexData> scale_color(LVecBase4 const &color_scale) const
ConstPointerTo<GeomVertexData> scale_color(LVecBase4 const &color_scale, int num_components, GeomEnums::NumericType numeric_type, GeomEnums::Contents contents) const

Returns a new GeomVertexData object with the color table modified in-place to apply the indicated scale.

If the vertex data does not include a color column, a new one will not be added.

Returns a new GeomVertexData object with the color table replaced with a new color table that has been scaled by the indicated value. The new color table will be added as a new array; if the old color table was interleaved with a previous array, the previous array will not be repacked.

void set_array(std::size_t i, GeomVertexArrayData const *array)

Replaces the indicated vertex data array with a completely new array. You should be careful that the new array has the same length and format as the old one, unless you know what you are doing.

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

ConstPointerTo<GeomVertexData> set_color(LColor const &color) const
ConstPointerTo<GeomVertexData> set_color(LColor const &color, int num_components, GeomEnums::NumericType numeric_type, GeomEnums::Contents contents) const

Returns a new GeomVertexData object with the color data modified in-place with the new value.

If the vertex data does not include a color column, a new one will not be added.

Returns a new GeomVertexData object with the color table replaced with a new color table for which each vertex has the indicated value. The new color table will be added as a new array; if the old color table was interleaved with a previous array, the previous array will not be repacked.

void set_format(GeomVertexFormat const *format)

Changes the format of the vertex data. If the data is not empty, this will implicitly change every row to match the new format.

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

void set_name(std::string const &name)

Changes the name of the vertex data. This name is reported on the PStats graph for vertex computations.

bool set_num_rows(int n)

Sets the length of the array to n rows in all of the various arrays (presumably by adding rows).

The new vertex data is initialized to 0, except for the “color” column, which is initialized to (1, 1, 1, 1).

The return value is true if the number of rows was changed, false if the object already contained n rows (or if there was some error).

This can be used when you know exactly how many rows you will be needing. It is faster than reserve_num_rows(). Also see unclean_set_num_rows() if you are planning to fill in all the data yourself.

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

void set_slider_table(SliderTable const *table)

Replaces the SliderTable on this vertex data with the indicated table. There should be an entry in this table for each kind of morph offset defined in the vertex data.

The SliderTable object must have been registered prior to setting it on the GeomVertexData.

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

void set_transform_blend_table(TransformBlendTable const *table)

Replaces the TransformBlendTable on this vertex data with the indicated table. The length of this table should be consistent with the maximum table index assigned to the vertices under the “transform_blend” name.

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

void set_transform_table(TransformTable const *table)

Replaces the TransformTable on this vertex data with the indicated table. The length of this table should be consistent with the maximum table index assigned to the vertices under the “transform_index” name.

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

void set_usage_hint(GeomEnums::UsageHint usage_hint)

Changes the UsageHint hint for this vertex data, and for all of the arrays that share this data. See get_usage_hint().

Don’t call this in a downstream thread unless you don’t mind it blowing away other changes you might have recently made in an upstream thread.

void transform_vertices(LMatrix4 const &mat)
void transform_vertices(LMatrix4 const &mat, int begin_row, int end_row)
void transform_vertices(LMatrix4 const &mat, SparseArray const &rows)

Applies the indicated transform matrix to all of the vertices in the GeomVertexData. The transform is applied to all “point” and “vector” type columns described in the format.

Applies the indicated transform matrix to all of the vertices from begin_row up to but not including end_row. The transform is applied to all “point” and “vector” type columns described in the format.

Applies the indicated transform matrix to all of the vertices mentioned in the sparse array. The transform is applied to all “point” and “vector” type columns described in the format.

void unclean_set_format(GeomVertexFormat const *format)

Changes the format of the vertex data, without reformatting the data to match. The data is exactly the same after this operation, but will be reinterpreted according to the new format. This assumes that the new format is fundamentally compatible with the old format; in particular, it must have the same number of arrays with the same stride in each one. No checking is performed that the data remains sensible.

bool unclean_set_num_rows(int n)

This method behaves like set_num_rows(), except the new data is not initialized. Furthermore, after this call, any of the data in the GeomVertexData may be uninitialized, including the earlier rows.

This is intended for applications that are about to completely fill the GeomVertexData with new data anyway; it provides a tiny performance boost over set_num_rows().

This can be used when you know exactly how many rows you will be needing. It is faster than reserve_num_rows().

bool validate_ptr(void const *ptr)
void write(std::ostream &out, int indent_level = 0) const