On 19/01/16 12:15, Osma Suominen wrote:
18.01.2016, 16:21, Andy Seaborne wrote:
Should be fixed now. I must have left some maven-generated files around.
Thanks Andy, it works for me now!
PS What I don't understand is why the SDB command line tools are so slow
to start up.
#define slow?
For me, sdbinfo is about 1.5 seconds:
$ time bin/sdbinfo --version
Jena: VERSION: 3.1.0-SNAPSHOT
Jena: BUILD_DATE: 2016-01-18T22:45:03+0000
ARQ: VERSION: 3.1.0-SNAPSHOT
ARQ: BUILD_DATE: 2016-01-18T22:45:03+0000
RIOT: VERSION: 3.1.0-SNAPSHOT
RIOT: BUILD_DATE: 2016-01-18T22:45:03+0000
SDB: VERSION: 3.1.0-SNAPSHOT
SDB: BUILD_DATE: 2016-01-19T14:05:40+0200
real 0m0.989s
user 0m1.560s
sys 0m0.096s
My guess is that this is because sdbinfo (and others) calls sdb_init,
which in turn calls sdb_path, which calls make_classpath (or
make_classpath_mvn). All those bash startups add up.
I can live with a couple of seconds startup delay, but refactoring the
scripts is an option.
Ping - light bulb moment. I have scripts of greater complexity and they
fly.
It's the make_classpath_mvn when using dev classes which calls maven to
make the classpath. And maven can go out to the web ...
It's an old way of doing it - Jena's newer scripts are much more slick
and my "set dev" one is similarly much newer. They do "<last
version>/lib/*" to get the dependencies then prepend target/classes for
most of -Pdev modules.
And can run different versions which helps for questions.
I'll put them ("it"; one script, many names) somewhere if people would
find it useful. And can improve them - I'm not claiming its great bash
scripting. Maybe there are better ways people are doing it.
Andy
-Osma
