I seriously want to start this relationship with you, the readership with pure honesty: I completely agree with the incredibility of C. However, before we get stuck into the detail, do you have any idea about C extension and why you should write one in 2016?<\/p>\n
Here\u2019s a discussion of a real-world need for a C extension in the Ruby language and how a non-C developer can get it done right.<\/p>\n
There are two reasons behind it. One of them is speed and another is leverage. If there is a well defined problem and if there is a need for speed, then writing an extension can be a way to go. <\/p>\n
Sharing my personal experience, writing C extension is a tough job and it an it advisable to go with FFI \u2013 don\u2019t use it, until you have no other option. But, frankly speaking, I was scared of writing C into widely used application.<\/p>\n
Apart from FFI, there were great options like for writing Ruby Extensions in Go, Crystal, Rust and many other languages where C bindings can be exposed. Anyways All I wanted to do is Go with C extension.<\/p>\n
- rtosc<\/code><\/h2>\nConsidering several OSC and canonical implementations, it made me use macros and structs,which I didn’t completely understand at the time. I was about to give up and fortunately I stumbled on the rtosc<\/code> library. It is consisted with two files rtosc.h<\/code> and rtosc.c<\/code>. Wishing the best spirit of open source, I enquired on GitHub issue and within few hours got a thoughtful and polite response -viva la internet!<\/p>\nThis helped me to try writing a C extension and trust me I’m glad that I did.<\/p>\n
Source Code – https:\/\/www.xavierriley.co.uk\/writing-a-c-extension-for-ruby-in-2016\/”sif”, [“This is a string”, 123, 3.14]<\/code><\/p>\nSo here you can have a look at the type of the arguments corresponds to a tag – s for a string, i for an int and for the Ruby API I decided to hide this so you do not consider about types.<\/p>\n
\r\n>> FastOsc.encode_single_message(\"\/foo\", [\"baz\", 1, 2.0])\r\n=> \"\/foo\\x00\\x00\\x00\\x00,sif\\x00\\x00\\x00\\x00baz\\x00\\x00\\x00\\x00\\x01@\\x00\\x00\\x00\"\r\n<\/pre>\nIt\u2019s a tag string and an output array of the encoded versions.<\/p>\n
Partner with us to turbocharge your entrepreneurial journey with a master-blaster Rails Mobile App.<\/span>
\nThe ideal Ruby on Rails Development Company<\/a> Awaits you!<\/i><\/strong><\/p>\nRuby’s C API<\/h2>\n
\nWorking out the type<\/strong><\/p>\n\r\nswitch(TYPE(current_arg))\r\n<\/pre>\nInts and Longs<\/strong><\/p>\n\r\ncase T_FIXNUM\r\n<\/pre>\nFloats<\/strong><\/p>\n\r\ncase T_FLOAT:\r\n<\/pre>\nTo convert from Ruby to C NUM2DBL<\/p>\n
Strings and Symbols<\/strong><\/p>\n\r\ncase T_STRING:\r\n<\/pre>\nTo convert Ruby symbols use rb_sym_to_s()<\/code>.<\/p>\nUnicode Strings<\/h2>\n
I had broken Unicode strings. However it needs to be fixed so did my job this way.<\/p>\n
\r\nstring_arg = rb_str_new2(string_from_c); \r\n\r\nenc = rb_enc_find_index(\"UTF-8\"); \r\n\r\nrb_enc_associate_index(string_arg, enc); \r\n<\/pre>\nRuby Time<\/code> objects<\/strong><\/p>\n Start with the object type<\/strong><\/p>\n\r\ncase T_DATA:\r\n<\/pre>\nIt means you have \tRuby object and you can specifically check using<\/p>\n
if (CLASS_OF(current_arg) == rb_cTime)<\/pre>\nOSC Timestamps (NTP)<\/h2>\n\r\n#define JAN_1970 2208988800.0 \/* 2208988800 time from 1900 to 1970 in seconds *\/\r\n\r\nuint64_t ruby_time_to_osc_timetag(VALUE rubytime) { \r\n uint64_t timetag;\r\n double floattime;\r\n uint32_t sec;\r\n uint32_t frac;\r\n\r\n switch(TYPE(rubytime)) {\r\n case T_NIL:\r\n timetag = 1;\r\n break;\r\n default:\r\n \/\/ convert Time object to ntp\r\n floattime = JAN_1970 + NUM2DBL(rb_funcall(rubytime, rb_intern(\"to_f\"), 0));\r\n\r\n sec = floor(floattime);\r\n frac = (uint32_t)(fmod(floattime, 1.0) * 4294967296); \/\/ * (2 ** 32)\r\n \/\/ printf(\"\\nsec: %04x\\n\", sec);\r\n \/\/ printf(\"\\nfrac: %04x\\n\", frac);\r\n timetag = (uint64_t)((uint64_t)sec << 32 | (uint64_t)frac);\r\n \/\/ printf(\"\\ntimetag: %08llx\\n\", timetag);\r\n break;\r\n }\r\n\r\n return timetag;\r\n}\r\n<\/pre>\nfurther reading<\/h2>\n\r\nEncoding \r\n-------------------------------------------------\r\n fast_osc 909.680k (\u00b121.4%) i\/s - 4.328M\r\n\r\n samsosc 271.908k (\u00b119.3%) i\/s - 1.327M\r\n\r\n oscruby 94.678k (\u00b114.5%) i\/s - 468.968k\r\n\r\nDecoding \r\n-------------------------------------------------\r\n fast_osc 2.635M (\u00b122.2%) i\/s - 12.435M\r\n\r\n samsosc 264.614k (\u00b116.1%) i\/s - 1.304M\r\n\r\n oscruby 36.362k (\u00b116.3%) i\/s - 179.622k\r\n<\/pre>\noscruby<\/code> non-optimised library for Ruby.<\/h2>\nsamosc<\/code> is hand rolled optimization for pure Ruby version and fast_osc is this C extension. So it is like 4x improvement in Encoding to OSC and a 10x improvement in Decoding over the fastest pure Ruby implementation. It was done completely in a successful way.<\/p>\nThe code is available here: https:\/\/github.com\/xavriley\/fast_osc<\/a>.<\/p>\n