catwell / luajit-msgpack-pure Goto Github PK

2018/03/02 19:13:45 [error] 31794#0: *287 lua entry thread aborted: runtime error: /data/web/openresty-websocket/resty/luajit-msgpack-pure.lua:433: unimplemented
stack traceback:
coroutine 0:
[C]: in function 'error'
/data/web/openresty-websocket/resty/luajit-msgpack-pure.lua:433: in function 'dynamic'
/data/web/openresty-websocket/resty/luajit-msgpack-pure.lua:567: in function 'unpack'
/data/web/openresty-websocket/libs/pushs.lua:77: in function 'unpack'
/data/web/openresty-websocket/websocket.lua:180: in function 'run'
content_by_lua(nginx.conf:106):3: in function <content_by_lua(nginx.conf:106):1>, client: 116.115.11.21, server: ngx.p.qc.com, request: "GET /?order=ccf8f35a2bff&user=2d7zsk&ver=372&reconnect=1&ingame=0 HTTP/1.1", host: "api2.qc.com:4396"
`

nginx_error. log a lot of errors. What should I do?

I maintain a pure javascript implementation of msgpack for node.js and the browser. I've extended the msgpack format to allow the js type undefined as well as binary blobs (since js strings are unicode and can't hold arbitrary binary data).

I noticed that you extended the format to support nan, inf, and -inf. Since JavaScript also has these values, (and I want to use your library for my luvit project) I propose we merge our extensions to become compatible.

What do you think?

Rob Pike says checking for endianness is wrong and I tend to agree. Moreover I fear my code doesn't work on big endian platforms (it's never been tested).

http://commandcenter.blogspot.fr/2012/04/byte-order-fallacy.html

Otherwise packing this would be disastrous:

{
  [0] = 1,
  [1000000] = 1,
}

The heuristic should be: integer keys are continuous and start at one.
Moreover current code is probably broken for arrays that start at 0 or negative numbers.

for a better time resolution in benchmark, use require'socket'.gettime() instead of require'os'.clock()

I don't know if this would work with luaffi -> try it someday.

A patch for convenient reuse of a buffer, and concatenating msgs without .. on many strings.

--packer functions
local pack_meta
pack_meta.__index = pack_meta
-- pack more stuff to the buffer
function p:pack(data)
    buffer = self
    packers.dynamic(data)
end
-- reuse the buffer for a new set of batched msg
function p:reset()
    self.size = 0
end

local function get_pack_buffer()
    local tbl = {}
    sbuffer_init(tbl)
    ffi.gc(tbl.data, ffi.C.free)
    return setmetatable(tbl,pack_meta)
end

return {
    pack_many = get_pack_buffer,
    pack = ljp_pack,
    unpack = ljp_unpack,
    set_fp_type = set_fp_type,
    set_table_classifier = set_table_classifier,
    packers = packers,
    unpackers = unpackers,
}

msgpack ruby gem was not using str 8 at all up to version 0.5, but since 0.6 it has started to use it. New users will get a failure.

msgpack 0.6:

2.1.2 :004 > MessagePack.pack(' ' * 255)
=> "\xD9\xFF "
msgpack 0.5.11:

2.1.2 :002 > MessagePack.pack(' ' * 255)
=> "\xDA\x00\xFF

Currently numbers are serialized as doubles. There should be an option to serialize them as floats to save memory at the expense of precision.