Avoid multiple redefinition when including codecfactory.h from different .cpp files

[pps83]

fixes #83

[pps83]

@lemire I have a question. Your github seem to have multiple int compression libs: https://github.com/lemire/SIMDCompressionAndIntersection https://github.com/lemire/streamvbyte https://github.com/lemire/simdcomp

What's the relationship? It seems that fastpfor includes streamvbyte (is it older version, does it need to be updated with upstream?). What about simdcomp vs fastpror?

[lemire]

This PR would effectively entice users to use the same codec instances throughout their application, and in particular across different threads. This goes contrary to our documentation which states that users should have different codecs per threads to avoid data races.

[lemire]

Your github seem to have multiple int compression libs:

Indeed. If you find that the documentation of one of them, or of many of them, is inadequate, please open an issue specific to that library explaining why the documentation is inadequate, and ideally, making proposal regarding how the documentation of that library be improved.

[lemire]

If you have multiple definitions, then your code is almost surely unsafe.

[lemire]

We are going to go with a simpler approach that should be safer: https://github.com/lemire/FastPFor/pull/92

[pps83]

We are going to go with a simpler approach that should be safer: #92

In my own project I have wrappers that use these IntegerCODECs as temporary vars:

void compressStreamvbyte(const uint32_t* in, const size_t length, uint32_t* out, size_t& nvalue)
{
    StreamVByte().encodeArray(in, length, out, nvalue);
}

Most coders don't have any state, so, this should be ok.

[lemire]

Most coders don't have any state, so, this should be ok

I agree.