Performance misleading - no JSON with gzip

[ChrisJefferson]

Why do you show vpack with gzip, but not JSON with gzip (especially as JSON tends to gzip very well)

[ChrisJefferson]

I took the first file, api-docs.json, and with gzip -9, json wins! (size 99,932)

[neunhoef]

Originally, we did not have compression in this table at all, because one of the most important features of VelocyPack for us is that one can access substructures efficiently without deserialising! Obviously, this does not work with compression. We have added the gzipped sizes of the compact VelocyPack because somebody asked for them.

You are right, we should also add a column for the sizes of the gzipped JSON documents. I would not be surprised if they are comparable or even smaller to the sizes of the compressed VelocyPack, since the "overhead" of JSON can be zipped relatively well. VelocyPack adds binary index tables which can probably not be zipped very well.

I'll keep this open until we got round to add the numbers.

[ChrisJefferson]

I should have said, for me you could also remove the gzips (or perhaps put them in another table with an explanation).

P.S. Hi @neunhoef

[neunhoef]

A different table would be an alternative approach, removing would probably make the person who asked for it unhappy.

P.S. Hi @ChrisJefferson