On 01/12/14 13:17, Kristian Rosenvold wrote:
I'd be interested in patching jena to use the thread-safe
FastDateFormat of commons-lang3 3.3.2 (or 2.6). The current use of the
non-threadsafe SimpleDateFormat class is a bit of a performance slob,
and
FastDateFormat is designed as direct replacment.
This would add commons-lang3 as a dependency some new places. Is is
worth spending my time on this ?
Kristian
Sounds great.
commons-lang3 is already a recursive dependency (via thrift) and gets
shipped in the distribution binaries so it's just a matter of adding it
explicitly to jena-parent for version and jena-core for the dependency.
[INFO] org.apache.jena:jena-arq:jar:2.12.2-SNAPSHOT
[INFO] +- org.apache.thrift:libthrift:jar:0.9.1:compile
[INFO] | \- org.apache.commons:commons-lang3:jar:3.1:compile
I grepped for SimpleDataTime in some of the main modules and found
RequestLog.java
jena-fuseki2/src/main/java/org/apache/jena/fuseki/server
schemagen.java
jena-core/src/main/java/jena
StringUtils.java
jena-arq/src/main/java/com/hp/hpl/jena/sparql/util
Utils.java
jena-arq/src/main/java/com/hp/hpl/jena/sparql/util
The speed ups due to special integer to StringBuffer is not new -
DateTimeStruct does similar. At the time I thought it was a local issue,
maybe not.
Does this change at Java8 and DateTimeFormatter? I haven't tried that
at all. (This is isn't a reason to not do it - it may be a reason to
hide things behind local libraries)
Andy
