Upstream changes:
Many performance improvements
Error message improvments
Lots of refactoring
Most of the JQValue interface changes in gojq fork had to be reworked but
resultet in a much nicer and cleaner changeset.
fq changes:
Assignment to JQValue (like a decode value) now shallowly converts the value into
a jq value before assigning. Was a bit hacky as it was and this makes JQValue
behave more like real jq values. This also fixes some advanced indexing issues.
Actual custom path updates will be something for the future.
Refactor to use groups between network layers to make them less coupled and reusable:
link_frame (ethernet etc)
inet_packet (ipv4,ipv6 etc)
ip_packet (tcp packet (not stream), udp segment etc)
Rename data to payload as i think it makes more sense for network data
Padding could end up double.
Remove to*range($pad) as it probably just confusing to be able
to pad an existing range, what to show in hexdump etc? zero bits
that do not actuall exist at that range?
Add tests and binary stdout support to tests
repl now rewrites query to do implicit display inside the sub eval.
This makes it possible to interrupt eval and output in a better and faster way.
Make JSON encoder fail early on errors.
Add more interrupt tests.
Add explode for binary that explode into an array of the binary unit tobits
Remove scan_toend as it feels a bit weird and one can use tobytes | .[match(...):] instead
Some general cleanup also
Now repl, slurp and help implemented using same query rewrite.
Include filename context in error if possible.
Add spew function that does opposite of slurp.
Start of help infra, not done or documented yet.
Show error pointer on parse error.
Rename internal eval to _eval and make eval be wrapper that
does rewrite and has various eror handling etc.
Nicer repl, slupr and help errors.
Speeds up interp quite a bit.
Also good as i've start to use more includes to make code easier to follow
where thigns come from and also makes jq-lsp happier.
Also rename evalContext to evalInstace to make it less confused with context.
Rewrite non-sub-repl queries as "q" -> ".[] | q" and pass in inputs as array.
Before each input did an eval.
Also fixup error message a bit, skip line:column when they dont make sense.
Fixes missing function errors for fq builtins
Add includes to fix some other missing function errors
Add COMPLETION_TIMEOUT env to increase comepletion timeout during test using -race
There is also toradix($base; $table)/fromradix($base; $table) for custom symbols and bases.
Add radix tests
Part of cleaning up all convert/tranformation functions into to/from*.
Remove number_to_bytes as with better binary support it can be written as:
1234 | tobytes | [.[range(.size)]] and soon as 1234 | tobytes | explode
Rename buffer to binary. Still some work left what to call buffer/binary in decode code.
Document decode value and binary type
Fix proper unit padding for tobytes and add still undocumenated extra padding argument.
Add some additional binary tests
Basic support, can also do CER and DER but without any extra validation.
No schema support.
Redo format doc.md usage a bit, now format/<dir>/<format>.md instead.
Related to #20
Remove bitio.Buffer layer. bitio.Buffer was a kitchen sink layer with helpers
now it's just a buffer and most functions have been moved to decode instead.
bitio package now only have primitive types and functions simialar to standard
library io and bytes packages.
Make nearly eveything internally use bitio.Bit* interfaces so that slicing work
correctly this will also make it possible to start experimenting with more
complicated silcing helpers, ex things like:
breplace(.header.bitrate; 123) to get a new buffer with bitrate changed.
