Sebastien Ponce authored 6 months ago and Marco Clemencic committed 5 months ago

the column has also been changed to width 10 by default and fmt was used to implement a fixed width integer representation.
In practice the integer is printed normally up to 10 digits, then a scientific notation is used.
Here are examples of the output :
  32401234
 324012345
3240123456
3.2401e+10
3.2401e+11

Sebastien Ponce authored 6 months ago and Marco Clemencic committed 5 months ago

Basically recreate the output of the GaudiHistoAlg (now gone)
This output is then understood by the testing machinary and will be checked automatically versus reference

This reverts commit a9a4cad9.

It has been reported that the method sortedEntitiesAsJSON was using a lot of memory
as it was converting all entities to JSON in one go (keep in mind that an entity can
be a huge histogram and that there may be a lot of them).
It has thus been replaced by method `applyToAllSortedEntities` which applies a lambda
to all entities but respecting the order that was there in sortedEntitiesAsJSON.
At any given moment, only one entity is converted to JSON in memory, improving
a lot memory usage.

Sebastien Ponce authored 1 year ago and Marco Clemencic committed 1 year ago

The problem was due to the use of deque, extremely slow for deleting when getting big.
The code was written that way supposing we would get ~100 entities (aka some human readable quantity), but we seem to reach > 100000...
So deque has been replaced by a map with entity's memory address as a key, and now the deletion is fast

Sebastien Ponce authored 2 years ago and Marco Clemencic committed 1 year ago

Also implemented a default, empty reset.

Sebastien Ponce authored 2 years ago and Marco Clemencic committed 1 year ago

The Sinks have now to implement a flush method, called both at the end (finalize) and regularly if auto flush is activated (via non 0 AutoFlushPeriod property).
Existing Sinks have been adapted.

Also added unit testing for BaseSink, both for the regular flushing and previous existing (and not yet tested) features.

Makes other sinks simpler and allows every sink to benefit from regexp based selection of entities

This allows to remove entities from the hub (and thus from the sinks in
the back) and thus allows to use entities in temporary objects, and not
only in Tool/algorithms.

The counter part is that BufferableCounters are no more movable,
although I doubt anyone really want to move a counter. It only has the
consequence that you can hardly have vectors of counters, but arrays or
deques are still fine

so for the type "category:subtype:precision" we look in the registry of
formats for "category:subtype:precision", "category:subtype" and
"category"

includes :
  - change of names of the classes, dropping the 'Counter' part which is
    an implementation detail
  - rewrote Root sink in a more generic and cleaner way
  - adapted json representation of histograms to changes in the new
    monitoring merge request

Conflicts:
	GaudiCoreSvc/src/MessageSvc/MessageSvcSink.cpp

Gitlab CI authored 4 years ago and Sebastien Ponce committed 4 years ago

patch generated by https://gitlab.cern.ch/lhcb/Gaudi/-/jobs/10087220

Note that MonitoringHub is now Gaudi::Monitoring::Hub.

Counters are now serialized in json and sent to a MessageSvcSink for printing.
This should be kind of backward compatible with current printing except that the Counter order may change.

The CounterHolder wrapper, now useless, has been dropped and counters register directly to the Monitoring::Hub
Note that his had the consequence that the Counters method of the python interface had to be dropped.
It does not seem to be used, but in case it would be needed, it can be implemented if we add a dedicated method to list counters of a given component to the Monitoringi::Hub