I found a couple of adjacent weird things:
There are a bunch of places in the json code that use
appendBinaryStringInfo() where appendStringInfoString() could be used, e.g.,
appendBinaryStringInfo(buf, ".size()", 7);
Is there a reason for this? Are we that stretched for performance? I
find this kind of code very fragile.
Also, the argument type of appendBinaryStringInfo() is char *. There is
some code that uses this function to assemble some kind of packed binary
layout, which requires a bunch of casts because of this. I think
functions taking binary data plus length should take void * instead,
like memcpy() for example.
Attached are two patches that illustrate these issues and show proposed
changes.
