[6.x.x] New exist:time XQuery Pragma

[adamretter]

Backported from FusionDB - A new XQuery Pragma called exist:time that is backwards compatible with the previous limited exist:timer pragma.

Like its predecessor exist:timer it can be used for timing an XQuery expression. However, its main advantage is that it can also be used to log multiple iterations of the same expression using the parameter measurement-mode=multiple. In addition to that the new exist:time pragma also adds:

1. Nanosecond precision (useful for measuring very fast expressions or multiple iterations thereof)
2. A humane format output of the total time elapsed.
3. Automatically adapts output between ns and ms units depending on the magnitude of the measurement
4. A logging-level parameter so that you may set the log-level dynamically on invocation.
5. Alog-message-prefix so that you may specify a prefix to be appended to the log message; this can make it much easier for you when trying to locate specific log messages.

Example 1 - Single measurement

(# exist:time logging-level=INFO #) {
    doc("/db/hindawi-example.xml")//p
}

produces log output like:

Elapsed: 21 ms.

Example 2 - Single iteration measurements inside FLWOR

for $i in (1 to 2)
return
  (# exist:time logging-level=INFO #) {
      $i * 10
  }

produces log output like:

Elapsed: 35967 ns.
Elapsed: 18028 ns.

Example 3 - Multiple iteration measurement inside FLWOR

This new facility is extremely useful for helping diagnose performance issues when you have a loop that is slow, but the expressions inside the loop itself appear fast. In such a scenario you can use exist:time to measure the performance of one or more expressions within the loop.

for $i in (1 to 2)
return
  (# exist:time logging-level=INFO measurement-mode=multiple #) {
      $i * 10
  }

produces log output like:

Elapsed: 53995 ns (53995 ns) [iterations=2 first=35967 ns, min=18028 ns, avg=26997.50 ns, max=35967 ns, last=18028 ns].

[dizzzz]

builds are not clean, rebase needed?

[adamretter]

@dizzzz The build errors are unrelated, they are from downloading the NVE CVE database.