Pervasives's deprecation (and thus provide OCaml 5.00 support).bigarray's ocamlfind package (and thus provide OCaml 5.00 support).Box{1,2,3}.inset. Rather than return the empty box when the size in a dimension i become negative, clamp it to 0 and use the ith coordinate of the mid point of the box for the ith coordinate of the resulting box's origin. This means that insetting boxes with large values eventually degenerates to the mid point of a box instead of the empty box. This avoids losing a box's location when one grows and shrinks them arbitrarily, e.g. in reaction to user input. Thanks to Michel Schinz for suggesting this better semantics.Gg.Float.pp hexadecimal notation renderer to use the built-in "%h" string introduced in OCaml 4.03.0. Nans, zeros and infinities will render differently. Use the deprecated Gg.Float.pp_legacy if you need to recover the old hex rendering.The Gg.Float module now includes Stdlib.Float (#19). Some values initially implemented in Gg.Float now use Stdlib.Float's definition or are deprecated in favour of corresponding functionality named differently. Implementations may differ but this shouldn't matter most of the time except for the first three items in this list:
Gg.Float.equal is deleted in favour of Stdlib.Float.equal The implemention differs, it moves from x = y to compare x y = 0 which differs on nan values. Stdlib.Float.equal treats them as equal Gg.Float.equal does not.Gg.Float.round is deleted and becomes Stdlib.Float.round. The implementation and behaviour on negative numbers differs. Gg.Float.round always rounded towards positive infinity on ties (-2. on -2.5). Stdlib.Float.rounds away from zero on ties (-3. on -2.5).Gg.Float.round_to_int is affected by the new round implementation (see previous point).Gg.Float.compare is deleted and becomes Stdlib.Float.compare (same implementation).Gg.Float.pi is deleted and becomes Stdlib.Float.pi, the bit pattern of the value is unchanged.Gg.Float.is_inf is implemented by Stdlib.Float.is_infinite and deprecated in favour of it (different implementation).Gg.Float.is_int is implemented by Stdlib.Float.is_integer and deprecated in favour of it (different implementation).Gg.Float.is_nan is deleted and becomes Stdlib.Float.is_nan (same implementation)Gg.Float.fmax is implemented by Stdlib.Float.max_num and deprecated in favour of it. The result of Gg.Float.fmax (-0.) (+0.) is changed, it returns +0. instead of -0..Gg.Float.fmin is implemented by Stdlib.Float.min_num and deprecated in favour of it. The result of Gg.Float.fmin (+0.) (-0.) is changed, it returns -0. instead of +0..Gg.Float.sign_bit is deleted and becomes Stdlib.Float.sign_bit (different implementation).Gg.Float.succ is deleted and becomes Stdlib.Float.succ (different implementation).Gg.Float.pred is deleted and becomes Stdlib.Float.pred (different implementation).Gg.Float.nan is renamed to Gg.Float.nan_with_payload to leave room for Stdlib.Float.nan's constant.Color.to_srgbi (inverse of Color.v_srgbi). Thanks to Christophe Troestler for the patch.Float.{is_nan,equal,compare}. Polymorphic equality was being used. Thanks to Christophe Troestler for the report (#17).Gg.M3.rot2 ?pt:(Some _) (#18).Box{1,2,3}.add_pt. Thanks to Christophe Troestler for the suggestion.V{2,3,4}.norm avoid {under,over}flows. Thanks to Christophe Troestler for the report and guidance.Size.of_w. Thanks to @rand00 for the report and the fix.Box1.pp and add to toplevel support.Box{1,2,3}.subset functions. Thanks to Armaël Guéneau for the report.#require "gg" no longer automatically opens Gg and installs printers. You now have to #require "gg.top" for this to happen.Gg.Ba (experimental). Convenience module for linear 1D bigarrays. The library now depends on Bigarrays.Gg.Raster module.to_string functions, they were not thread safe and we now have Format.asprintf which can be used with the pretty printers.M3.rot2 to specify a center for the rotation.M{2,3,4}.mul on M{2,3,4}.id arguments.M4.{move2,rot2,scale2,rigid2}M4.{ortho,persp}, replace ~bottom by ~bot.Size1 module for sizes in 1D space.Box1 module for 1D axis-aligned boxes (closed intervals).Size2.{aspect,of_h,of_w} functions.Box2.{bm_pt,ml_pt,mm_pt,mr_pt,tm_pt} for accessing middle points on the sides.Box2.{bottom,top}_{left,right} to Box2.{b,t}{l,r}_pt.Box3.{fbl,fbr,ftl,ftr,nbl,nbr,ntl,ntr} corner accessors.Box3.inset, new size was incorrectly computed.The following functions were renamed so that each module uses the same name for the same transform. Previously we had e.g. M3.scale and M4.scale3 (3D) and M4.scale (4D) which is confusing and inconvenient when one wants to switch from one matrix to the other.
M2.{rot,scale} to M2.{rot2,scale2}.M3.{move,rot,rigid,srigid} to M3.{move2,rot2,rigid2,srigid2}.M3.{rot_{map,axis,zyx},scale} to M3.{rot3_{map,axis,zyx},scale3}.M4.{move,rot_{map,axis,zyx},scale,rigid,rigidq,srigid,srigidq} to M3.{move3,rot3_{map,axis,zyx},scale,rigid3,rigid3q,srigid3,srigid3q}M4.scale to M4.scale4M4.{rot_{map,axis,zyx},to_rot_{axis,zyx} to M4.{rot3_{map,axis,zyx},to_rot3_{axis,zyx}.First release. Part of the work was sponsored by Citrix Systems R&D and OCaml Labs.