F1.PL README
============

The timing PDFs from the FIA are normally available very quickly after each
session but the links can be removed within a few days if there are back to
back races. It's a good idea to back up the PDFs as the database can
be re-created from them at any time, but as the document names will not
change from season to season there can be a danger of overwriting them.

When updating the database, if there are existing records for the source
documents they will be deleted first. This is done within a transactions so
the delete will only be carried out if all the new records are successfully
added.

The database isn't fully relational or normalized, e.g., the drivers' names
are included in the race_fastest_lap, as this minimizes the dependencies. Only
the documents of interest need be added to the database and this can be done
in any order. The driver tables for each practice, qualifying session and the
race are updated at the same time as the relevant lap-time table. To get all
the race lap times both the race-analysis and the race-history PDFs must be
used as the race-analysis has the start time instead of a lap time for the
first lap.  Use one of the race_lap SQL views to get all the lap times. 

Not all the PDFs can be processed;  the race grid and the race and qualifying
classification documents are updated with facsimile versions of the time sheets
signed by the stewards from which the text cannot be extracted. Also, unusual
race events may mean that the regular expressions used in the script to parse
the text may not work correctly or that the layout of the timing sheet is
compromised e.g., on lap 25 in the Race History Chart for the 2011 Canadian
Grand Prix there is insufficient space for both the lap time (which includes
the red flag period) and the gap.

SQLite3 has been used for the database as it's small and easy to install but
it is not ideal for accurate time arithmetic. Time is expressed as a decimal
value, e.g., 0.5 is 12 hours or midnight, so this together with rounding 
errors in computer floating number calculations means that there maybe small 
errors when summing lap times to thousandths of a second. It is possible to 
create an aggregate extension using the SQLite C API but this would need to be
compiled separately for each platform. Alternatively it would be possible to
up-size to another RDMS such as PostgreSQL which has better support for time
arithmetic.

There is a SQL query to calculate the cumulative difference between each
driver's lap time and the race winner's average lap time. It is only for one
race and is very slow so it would be better in a RDMS with stored procedures
and windowing functions, such as PostgreSQL 9. Alternatively further analysis
can be carried out with a spreadsheet using the export function of the Perl
script or from the Firefox add-in, SQLite Manager.