In order to calculate the efficiency of the axial tracking
trigger on events other than bhabhas (where you can use the
CC-based bhabha trigger) you have to do a bit of work
(since every other trigger line requires the axial trigger
to get the event).  The trigger line that requires the
fewest number of axial tracks is the ElTrack line, and this
trigger line is convieniently on in a moderate number of 
hadronic events and in most (all?) radiative bhabha events.

We would like to measure the efficiency of the axial trigger
to find "hadrons" (pions and kaons) at all momenta.  The 
prescription described below works fine for this, except 
when miniPass2 is not processing hadronic events (as it was
for most of the 4S running).  In this case, you can use
radiative bhabha events to sample of range of momenta.
All of the tracks in this case will be electrons (or 
positrons), but at least you sample a range of momenta.

Below is a description of what the code does:

First it checks the TriggerL1Data to see if the ElTrack line
was on.  If it was not on, the event is skipped.

Next it loops over all of the tracks, selecting the "good" ones.
Good is defined as not being an aborted fit, not having the track's
origin fixed to (0,0) due to insufficient number of stereo layers
(this can lead to a deficit of tracks at the largest abs(cos(theta))
values, however, we know from experience that the trigger gets more
efficient, if there is any change, at higher values of abs(cos(theta))),
having a momentum between 6.0 GeV and 100 MeV, having a hit fraction
of greater than 40%, having abs(cos(theta)) <= 0.93, having chi^2/degree
of freedom < 3.5 (this is fairly tight...), and having a d_0 of <
1.5 cm.

For all of the good tracks we save it transverse momentum (p_t), its
cos(theta) <===== is this used???????
and its phi position at layer 9 (we save which layer 9 wire the
track passes closest to).  We also record which track has the
highest p_t.

Next we check the tracking trigger timing bit to make sure it is
in a reasonable position (in the triggering bucket, or within
two after it).  If not, we skip the event.

Next we check that the track with the highest p_t (stiffest track)
matches a trigger track.  An offline track is matched to a trigger 
track if it is in the same phi bin, or in an adjacent phi bin, as 
the trigger track.  (Recall that there are 112 phi bins in the trigger.)

Now we loop over all of the good offline tracks.  We require that the
track be more than 5 phi bins away from the stiffest track (or we
go on to the next track).  We then check that this track can be
uniquely identified with set of trigger phi bins, so that we can
properly record the success or failure of the axial tracking at the
p_t of the current track.  An offline track can be uniquely identified
with a section of the trigger if none of the other (good) offline
tracks have a phi within 2 of this track.  If the track cannot
be uniquely identified with a portion of the trigger, we skip this
track.

Next we check if the track is isolated (meaning that it is unlikely
that any other track intersects the current track between layers
1 and 16).  To be isolated we require that no other (good) offline 
track be within 5 phi bins of the current track.  The efficiency is
studied separately for isolated and nonisolated (but uniquely identified)
tracks.

We check if there is a matching trigger track to the current track
and record the success or failure at the p_t of the current track in
the appropriate histogram.  We segregate the data based on the level 3
classification.  If level 3 says the event is a bhabha, gamma-gamma,
or radiative bhabha we call the track an electron.  If level 3 calls
the event a hadron we call the track a hadron.  For any other level 3
classification we call the track junk.

For offline tracks without matching trigger tracks, we check if
there were at least 6 of the first 8 layers used and at least 6 of
the second 8 layers used.  If not, we histogram it as a failure due
to missing DR hits.  Otherwise we histogram it as a failure due to
the trigger.

There are scripts to process the output (both from the log file 
and from the histogram file) in /home/trigger3/scripts/axtr.