Viewing and estimating with statistics

The statistics information collected for each table is easiest to see using the pg_stats view. The amount of information it returns is a bit overwhelming though, and not well suited to the standard display format. The following script is named table-stats.sh in the book's file set, and it tries to display the statistics information for a particular table (with an optional database too) in a way that makes the information as readable as possible:

#!/bin/bash
if [ -z "$1" ]; then
  echo "Usage: table-stats.sh table [db]"
  exit 1
fi
TABLENAME="$1"
if [ -n "$2" ] ; then
  DB="-d $2"
fi
PSQL="psql $DB -x -c "
$PSQL "
SELECT
  tablename,attname,null_frac,avg_width,n_distinct,correlation,
  most_common_vals,most_common_freqs,histogram_bounds
FROM pg_stats
WHERE tablename='$TABLENAME';
" | grep -v "\-\[ RECORD "

A typical usage would be:

./table-stats.sh...