thephoeron / bit-smasher Goto Github PK

Perhaps you could try to use ALEXANDRIA:DEFCONSTANT with :TEST #'EQUALP to turn this into a constant, in order to avoid (dynamic) variable lookup at runtime.

Building with SBCL 2.2.3 / ASDF 3.3.5 for quicklisp dist creation.

Trying to build commit id 78144ca

bit-smasher-test fails to build because of a failure in bit-smasher.

bit-smasher fails to build with the following error:

; caught ERROR:
;   READ error during COMPILE-FILE: Package BIT-SMASHER-ASD does not exist. Line: 29, Column: 75, File-Position: 969 Stream: #<SB-INT:FORM-TRACKING-STREAM for "file /home/quicklisp/quicklisp-controller/dist/build-cache/bit-smasher/be8e97d25483ed8f0d7bc470584ba5dc30ef356e/bit-smasher-20220707-git/src/packages.lisp" {1005F985C3}>
...
Unhandled UIOP/LISP-BUILD:COMPILE-FILE-ERROR in thread #<SB-THREAD:THREAD "main thread" RUNNING {1001BF0003}>: COMPILE-FILE-ERROR while compiling #<CL-SOURCE-FILE "bit-smasher" "packages">

Full log here

The BIT-SMASHER library would be more generally useful, and less naive, if it supported handling of signed-bytes, the option of converting between little- and big-endian, and direct bit-shifting implementations of all arithmetic operations.

There are a few libraries already available in Quicklisp that will make this enhancement more or less painless.

I used the library before the order parameter was removed from the bits->octets function and just maintained a local small patch to add it back - I can see how it's complicating the interface, so I could probably just use ironclad:integer-to-octets and supply the value myself. That said, the docstring still mentions the parameter.

Both performance and accuracy can be significantly improved by re-implementing the binary arithmetic algorithms directly using the new bit-ops library. This will also make it easier to work with signed integers, and other arbitrary binary structures, encoded as bit-vectors.

CC: @guicho271828