Drop rows where timestamps are older than subsequent row

public void onSensorChanged(SensorEvent event) {
  long systemTimeStamp= SystemClock.elapsedRealtimeNanos();
  double timeDifference = ((double) (systemTimeStamp - prevTime)) / 1000000d;    //msec        
  Log.e("No PROBLEM",timeDifference);
  prevTime=systemTimeStamp;
}

df.Time = pd.to_timedelta(df.Time)
df = df[df.Time == df.Time.max()]
print (df)
   ID Name     Time Comment
0   0  Foo 12:17:37    Rand
1   1  Foo 12:17:37   Rand1

df = df.groupby('Name', sort=False)
       .apply(lambda x: x[x.Time == x.Time.max()])
       .reset_index(drop=True)
print (df)
   ID Name     Time Comment
0   0  Foo 12:17:37    Rand
1   1  Foo 12:17:37   Rand1
2   4  Bar 09:01:00   Rand4
3   5  Bar 09:01:00   Rand5

c = ['Subject', 'Verb', 'Object']

def f(x):
    return x[c].duplicated() & x.Date.diff().dt.days.lt(5)

df = df.sort_values(c)
df[~df.groupby(c).apply(f).values]

  Subject     Verb Object       Date
0    Bill      Ate   Food 2015-07-11
1   Steve  Painted  House 2011-08-12
3   Steve  Painted  House 2011-08-25

> INSERT myMetric value1=1
> INSERT myMetric value1=10
> select value1 from myMetric
name: myMetric
time                value1
----                ------
1526032578114702408 1
1526034352621423990 10
> select difference(value1) from myMetric
name: myMetric
time                difference
----                ----------
1526034352621423990 9
> select derivative(value1,1s) from myMetric
name: myMetric
time                derivative
----                ----------
1526034352621423990 0.005071832014237941

select usr,asset
  , max(case event when 'out' then ts end) out_time
  , max(case event when 'in' then ts end) in_time
from (
  select f.*,
      -- first 'out' matches first 'in' and so on
      row_number() over(partition by usr, asset, event order by ts) as grp
  from foo f
  ) t
group by usr, asset, grp
order by out_time