DMelt:DataAnalysis/8 Metadata

Table of contents

Metadata

Metadata is a short record which captures main characteristics of an entire data record or objects inside each data record. These characteristics appear to be useful when one needs to find a necessary record as fast as possible, without reading the entire data records. Read Metadata article.

Let us consider a script which fills an event record with many events. It can look as

from jhplot.io import *
from jhplot import *

ps=math.Poisson(500)
def event(entry):
  p1,p2,p3=P0D('a'),P0D('b'),P0D('c')
  p1.randomUniform(ps.next(),0,1)
  p2.randomNormal(ps.next(),0,1)
  p3.randomNormal(ps.next(),0,1)
  return [str(entry),p1,p2,p3]

f=HFile('data.jser','w')
for i in range(5000):
  ev=event(i)
  if (i%100 == 0):
     print 'event=',ev[0]
  f.write(ev)
f.close()

In this example, event consists of 3 data containers with Poisson random numbers. We write all such events into a file. in a zipped serialized form. How to insert a data record characterizing the sum of 3 random values, x+y+x, such that after reading this small record one can navigate to the necessary event without running over all 50000 events?