# Transformations

These functions are declared in the main Allegro header file:

~~~~c
 #include <allegro5/allegro.h>
~~~~

Transformations allow you to transform the coordinates you use for drawing
operations without additional overhead. Scaling, rotating, translating, and
combinations of these are possible as well as using custom transformations.
There are two types of transformations that you can set, 'regular'
transformations and projection transformations. The projection transform is
rarely used in 2D games, but is common in 3D games to set up the projection
from the 3D world to the 2D screen. Typically, you would use the regular
transform for non-projective types of transformations (that is, translations,
rotations, scales, skews... i.e. transformations that are linear), while the
projection transform will be used for setting up perspective and possibly more
advanced effects. It is possible to do everything with just using the
projection transformation (that is, you'd compose the projection transformation
with the non-projection transformations that, e.g., move the camera in the
world), but it is more convenient to use both for two reasons:

- Regular transformations can be changed while the bitmap drawing is held (see
  [al_hold_bitmap_drawing]).

- Regular transformations work with memory bitmaps.

As a result, if you're making a 2D game, it's best to leave the projection
transformations at their default values.

Both types of transformations are set per target-bitmap, i.e. a change of the
target bitmap will also change the active transformation.

Allegro provides convenience functions to construct transformations in 2D and 3D
variants (the latter with a `_3d` suffix), so you don't have to deal with the
underlying matrix algebra yourself.

The transformations are combined in the order of the function invocations. Thus
to create a transformation that first rotates a point and then translates it,
you would (starting with an identity transformation) call [al_rotate_transform]
and then [al_translate_transform]. This approach is opposite of what OpenGL uses
but similar to what Direct3D uses.

For those who know the matrix algebra going behind the scenes, what the
transformation functions in Allegro do is "pre-multiply" the successive
transformations. So, for example, if you have code that does:

~~~~c
al_identity_transform(&T);

al_compose_transform(&T, &T1);
al_compose_transform(&T, &T2);
al_compose_transform(&T, &T3);
al_compose_transform(&T, &T4);
~~~~

The resultant matrix multiplication expression will look like this:

    T4 * T3 * T2 * T1

Since the point coordinate vector term will go on the right of that sequence of
factors, the transformation that is called first, will also be applied first.

This means if you have code like this:

~~~~c
al_identity_transform(&T1);
al_scale_transform(&T1, 2, 2);
al_identity_transform(&T2);
al_translate_transform(&T2, 100, 0);

al_identity_transform(&T);

al_compose_transform(&T, &T1);
al_compose_transform(&T, &T2);

al_use_transform(&T);
~~~~

it does exactly the same as:

~~~~c
al_identity_transform(&T);
al_scale_transform(&T, 2, 2);
al_translate_transform(&T, 100, 0);
al_use_transform(&T);
~~~~

## API: ALLEGRO_TRANSFORM

Defines the generic transformation type, a 4x4 matrix. 2D transforms use only
a small subsection of this matrix, namely the top left 2x2 matrix, and the
right most 2x1 matrix, for a total of 6 values.

*Fields:*

* m - A 4x4 float matrix

## API: al_copy_transform

Makes a copy of a transformation.

*Parameters:*

* dest - Source transformation
* src - Destination transformation

## API: al_use_transform

Sets the transformation to be used for the the drawing operations on the target
bitmap (each bitmap maintains its own transformation).
Every drawing operation after this call will be transformed using this
transformation. Call this function with an identity transformation to return
to the default behaviour.

This function does nothing if there is no target bitmap.

The parameter is passed by reference as an optimization to avoid the overhead of
stack copying. The reference will not be stored in the Allegro library so it is
safe to pass references to local variables.

~~~~c
void setup_my_transformation(void)
{
   ALLEGRO_TRANSFORM transform;
   al_translate_transform(&transform, 5, 10);
   al_use_transform(&transform);
}
~~~~

*Parameters:*

* trans - Transformation to use

See also: [al_get_current_transform], [al_transform_coordinates]

## API: al_get_current_transform

Returns the transformation of the current target bitmap, as set by
[al_use_transform].  If there is no target bitmap, this function returns NULL.

*Returns:*
A pointer to the current transformation.

See also: [al_get_current_projection_transform]

## API: al_use_projection_transform

Sets the projection transformation to be used for the the drawing operations on
the target bitmap (each bitmap maintains its own projection transformation).
Every drawing operation after this call will be transformed using this
transformation. To return default behavior, call this function with an
orthographic transform like so:

~~~~c
ALLEGRO_TRANSFORM trans;
al_identity_transform(&trans);
al_orthographic_transform(&trans, 0, 0, -1.0, al_get_bitmap_width(bitmap),
                          al_get_bitmap_height(bitmap), 1.0);

al_set_target_bitmap(bitmap);
al_use_projection_transform(&trans);
~~~~

The orthographic transformation above is the default projection transform.

This function does nothing if there is no target bitmap. This function also
does nothing if the bitmap is a memory bitmap (i.e. memory bitmaps always use
an orthographic transform like the snippet above). Note that the projection
transform will be reset to default if a video bitmap is converted to a memory
bitmap. Additionally, if the bitmap in question is the backbuffer, it's
projection transformation will be reset to default if it is resized. Lastly,
when you draw a memory bitmap to a video bitmap with a custom projection
transform, this transformation will be ignored (i.e. it'll be as if the
projection transform of the target bitmap was temporarily reset to default).

The parameter is passed by reference as an optimization to avoid the overhead of
stack copying. The reference will not be stored in the Allegro library so it is
safe to pass references to local variables.

Since: 5.1.9

See also: [al_get_current_projection_transform]

## API: al_get_current_projection_transform

If there is no target bitmap, this function returns NULL.

*Returns:*
A pointer to the current transformation.

Since: 5.1.9

See also: [al_use_projection_transform]

## API: al_get_current_inverse_transform

Returns the inverse of the current transformation of the target bitmap.
If there is no target bitmap, this function returns NULL.

This is similar to calling `al_invert_transform(al_get_current_transform())`
but the result of this function is cached.

> *Note*: Allegro's transformation inversion functions work correctly only 
with 2D transformations.

Since: 5.1.0

## API: al_invert_transform

Inverts the passed transformation. If the transformation is nearly singular
(close to not having an inverse) then the returned transformation may be
invalid. Use [al_check_inverse] to ascertain if the transformation has an
inverse before inverting it if you are in doubt.

*Parameters:*

* trans - Transformation to invert

> *Note*: Allegro's transformation inversion functions work correctly only 
with 2D transformations.

See also: [al_check_inverse]

## API: al_check_inverse

Checks if the transformation has an inverse using the supplied tolerance.
Tolerance should be a small value between 0 and 1, with 1e-7 being sufficient
for most applications.

In this function tolerance specifies how close the determinant can be to
0 (if the determinant is 0, the transformation has no inverse). Thus
the smaller the tolerance you specify, the "worse" transformations will
pass this test. Using a tolerance of 1e-7 will catch errors greater
than 1/1000's of a pixel, but let smaller errors pass. That means that
if you transformed a point by a transformation and then transformed it
again by the inverse transformation that passed this check, the resultant
point should less than 1/1000's of a pixel away from the original point.

Note that this check is superfluous most of the time if
you never touched the transformation matrix values yourself. The only
thing that would cause the transformation to not have an inverse is if
you applied a 0 (or very small) scale to the transformation or you have
a really large translation. As long as the scale is comfortably above 0,
the transformation will be invertible.

*Parameters:*

* trans - Transformation to check
* tol - Tolerance

*Returns:*
1 if the transformation is invertible, 0 otherwise

> *Note*: Allegro's transformation inversion functions work correctly only 
with 2D transformations.

See also: [al_invert_transform]

## API: al_identity_transform

Sets the transformation to be the identity transformation. This is the default
transformation. Use [al_use_transform] on an identity transformation to return
to the default.

~~~~c
ALLEGRO_TRANSFORM t;
al_identity_transform(&t);
al_use_transform(&t);
~~~~

*Parameters:*

* trans - Transformation to alter

See also: [al_translate_transform], [al_rotate_transform], [al_scale_transform]

## API: al_build_transform

Builds a transformation given some parameters. This call is equivalent to
calling the transformations in this order: make identity, rotate, scale,
translate. This method is faster, however, than actually calling those
functions.

*Parameters:*

* trans - Transformation to alter
* x, y - Translation
* sx, sy - Scale
* theta - Rotation angle in radians

> *Note*: this function was previously documented to be equivalent to a
different (and more useful) order of operations: identity, scale, rotate,
translate.

See also: [al_translate_transform], [al_rotate_transform],
[al_scale_transform], [al_compose_transform]

## API: al_build_camera_transform

Builds a transformation which can be used to transform 3D coordinates in world
space to camera space. This involves translation and a rotation. The function
expects three coordinate triplets: The camera's position, the position the
camera is looking at and an up vector. The up vector does not need to be of
unit length and also does not need to be perpendicular to the view direction -
it can usually just be the world up direction (most commonly 0/1/0).

For example:

~~~~c
al_build_camera_transform(&t,
    1, 1, 1,
    5, 5, 5,
    0, 1, 0);
~~~~

This create a transformation for a camera standing at 1/1/1 and looking towards
5/5/5.

> *Note*: If the *position* and *look* parameters are identical, or if the
*up* direction is parallel to the view direction, an identity matrix is created.

Another example which will simply re-create the identity matrix:

~~~~c
al_build_camera_transform(&t,
    0, 0, 0,
    0, 0, -1,
    0, 1, 0);
~~~~

An example where the up vector will cause the camera to lean (roll) by 45
degrees:

~~~~c
al_build_camera_transform(&t,
    1, 1, 1,
    5, 5, 5,
    1, 1, 0);
~~~~

Since 5.1.9

See also: [al_translate_transform_3d], [al_rotate_transform_3d],
[al_scale_transform_3d], [al_compose_transform], [al_use_transform]

## API: al_translate_transform

Apply a translation to a transformation.

*Parameters:*

* trans - Transformation to alter
* x, y - Translation

See also: [al_rotate_transform], [al_scale_transform], [al_build_transform]

## API: al_rotate_transform

Apply a rotation to a transformation.

*Parameters:*

* trans - Transformation to alter
* theta - Rotation angle in radians

See also: [al_translate_transform], [al_scale_transform], [al_build_transform]

## API: al_scale_transform

Apply a scale to a transformation.

*Parameters:*

* trans - Transformation to alter
* sx, sy - Scale

See also: [al_translate_transform], [al_rotate_transform], [al_build_transform]

## API: al_transform_coordinates

Transform a pair of coordinates.

*Parameters:*

* trans - Transformation to use
* x, y - Pointers to the coordinates

See also: [al_use_transform], [al_transform_coordinates_3d]

## API: al_transform_coordinates_3d

Transform x, y, z coordinates.

*Parameters:*

* trans - Transformation to use
* x, y, z - Pointers to the coordinates

Note: If you are using a projection transform you most likely will want
to use [al_transform_coordinates_3d_projective] instead.

Since 5.1.9

See also: [al_use_transform], [al_transform_coordinates]

## API: al_transform_coordinates_4d

Transform x, y, z, w coordinates.

*Parameters:*

* trans - Transformation to use
* x, y, z, w - Pointers to the coordinates

Since 5.2.4

See also: [al_use_transform], [al_transform_coordinates], [al_transform_coordinates_3d]

## API: al_transform_coordinates_3d_projective

Transform x, y, z as homogeneous coordinates. This is the same as using
[al_transform_coordinates_4d] with the w coordinate set to 1, then
dividing x, y, z by the resulting w. This will provide the same
coordinates Allegro will draw to when a projective transform is in
effect as set with [al_use_projection_transform].

*Parameters:*

* trans - Transformation to use
* x, y, z - Pointers to the coordinates

Since 5.2.4

See also: [al_use_transform], [al_transform_coordinates], [al_transform_coordinates_3d], [al_use_projection_transform]

## API: al_compose_transform

Compose (combine) two transformations by a matrix multiplication.

    trans := trans other

Note that the order of matrix multiplications is important. The effect
of applying the combined transform will be as if first applying `trans`
and then applying `other` and not the other way around.

*Parameters:*

* trans - Transformation to alter
* other - Transformation used to transform `trans`

See also: [al_translate_transform], [al_rotate_transform], [al_scale_transform]

## API: al_orthographic_transform

Combines the given transformation with an orthographic transformation which
maps the screen rectangle to the given left/top and right/bottom coordinates.

near/far is the z range, coordinates outside of that range will get clipped.
Normally -1/1 is fine because all 2D graphics will have a z coordinate of 0.
However if you for example do al_draw_rectangle(0, 0, 100, 100) and rotate
around the x axis ("towards the screen") make sure your z range allows values
from -100 to 100 or the rotated rectangle will get clipped.

Also, if you are using a depth buffer the z range decides the depth
resolution. For example if you have a 16 bit depth buffer there are only
65536 discrete depth values. So if your near/far is set to -1000000/1000000
most of the z positions would not result in separate depth values which
could lead to artifacts. 

Since: 5.1.3

See also: [al_use_projection_transform], [al_perspective_transform]

## API: al_perspective_transform

Like [al_orthographic_transform] but honors perspective. If everything is at
a z-position of -near it will look the same as with an orthographic
transformation.

To use a specific horizontal field of view you can use the relation:

    tan(hfov / 2) = (right - left) / 2 / near

Since: 5.1.3

See also: [al_use_projection_transform], [al_orthographic_transform]

## API: al_translate_transform_3d

Combines the given transformation with a transformation which translates
coordinates by the given vector.

Since: 5.1.3

See also: [al_use_projection_transform]

## API: al_scale_transform_3d

Combines the given transformation with a transformation which scales
coordinates by the given vector.

Since: 5.1.3

See also: [al_use_projection_transform]

## API: al_rotate_transform_3d

Combines the given transformation with a transformation which rotates
coordinates around the given vector by the given angle in radians.

Note: The vector is assumed to be of unit length (otherwise it will
also incur a scale).

Since: 5.1.3

## API: al_horizontal_shear_transform

Apply a horizontal shear to the transform

*Parameters:*

* trans - Transformation to alter
* theta - Rotation angle in radians

Since: 5.1.7

See also: [al_vertical_shear_transform]

## API: al_vertical_shear_transform

Apply a vertical shear to the transform

*Parameters:*

* trans - Transformation to alter
* theta - Rotation angle in radians

Since: 5.1.7

See also: [al_horizontal_shear_transform]
