Daniel Lemire

Very cool! What do you think about a short value optimization for strings 1-7 bytes inline? I would think an awful lot of keys are 7 bytes or less.

In a SIMD world, I'm not sure 7 bytes is a threshold I like all that much. I would imagine doing any optimization that kicks in and out at 7 bytes is a recipe for branch misses. We might regard a short string as 1-31 or 1-32 bytes.

"Just like the old Soviet-era joke about ‘mat’ (the amazing sub-dialect of Russian swearing) usage by workers in the missile factory, we have two stages, and one goes into the other." now, this is interesting! what is the full joke?

The joke was that that the CIA installed a listening device into a Soviet missile factory, and after 6 months of surveillance, concluded that: "The Soviet missile has two parts, and one goes into the other".

if i read this well, it does parse json in depth, it will not bork at any level of depth and creates a full tree this may be the parser i was waiting for, most json parsers give up somewhere

also, could this lib easily be modified to parse rss ? sorry, naieve and dumb question, trying to minimize my efforts

As far as I know this is equivalent to parsing XML. XML parsing is pretty well-worn. I don't know if the tricks we do would be directly applicable. The elements are much more verbose so the problem is a bit more like regex or string scanning (which I've done in the past).