#!/bin/bash
## invoked from WEBGETLOGS (cron-job) to do the daily webalizing -- Eugene Reimer 2009-12-08;
## receives DOM (domain-name), invokes webalize after determining which logfiles to pass;
## PREREQ:  dateplusdays -- from http://ereimer.net/programs/general-purpose-scripts.htm
## USAGE EXAMPLES:
##	webalize-auto  nativeorchid.org
##	webalize-auto  ereimer.net

if   [[ $@ == *debwendon*    ]];then	DOM=debwendon.org;     PROJ=debwendon;
elif [[ $@ == *ereimer*      ]];then	DOM=ereimer.net;       PROJ=er;
elif [[ $@ == *nativeorchid* ]];then	DOM=nativeorchid.org;  PROJ=noci;
fi
{
if ! [ $DOM ];then  echo "webalize-auto lacks support for names like $@";  exit 8;  fi			##quit if args invalid;  msg will get to tee??
DATE=$(date +%Y%m%d)											##get DATE in YYYYMMDD style
for W in /pix/WEBSTATS-$PROJ/wa-Monthly_*htm;do true; done						##get W newest webalizer-Monthly-report filename
WDAY=$(cat $W |tr '\n' ' '  |sed 's|</*TABLE|\n&|g; s|<TR|\n&|g' |					##convert to one-line-per-TR format
  sed -n '0,/Daily Statistics f/d; /NAME="HOURSTATS"/,$d; /<TR.*<TD/{s|</TD>.*||; s|<[^>]*>||g; p}' |	##keep only Daily-Stats rows, and only first column
  tail -n1)												##keep only the last of those
WYYYYMM=${W##*_};  WYYYYMM=${WYYYYMM%.htm}								##get year+month from $W
WDATEN=$WYYYYMM$(printf "%02d" $WDAY)									##get WDATEN last date in webalizer-report (first 2-hours)
WDATE=$(dateplusdays $WDATEN -1)									##get WDATE last full date webalized
ARGS=;													##init ARGS-list for webalize-call
for ((DT=WDATEN; DT<DATE; DT=$(dateplusdays $DT +1) ));do						##for each unwebalized day...
  F=/pix/WEBLOGS/access_log-$DOM.$DT									##set F to name of DAILY-logfile for $DT
  if ! [ -e $F ];then											##nonexistent-DAILY...
    if ((DT>=$(dateplusdays $DATE -6) ));then  echo "webalize-auto: quit at $DT";  break;  fi		##quit if only a few days ago (can fix), otherwise do best we can
  fi
  for H in $F-hourly*;do [ -e $H ] && ARGS="$ARGS $H"; done						##add all HOURLYs for $DT, if any, to ARGS
  [ -e $F ] && ARGS="$ARGS $F"										##add $F to ARGS
done
if [[ $ARGS != "" ]];then										##do nothing if ARGS empty
  echo "webalize-auto: calling webalize $ARGS"								##msg
  webalize $ARGS											##==webalize the ARGS-list
  WEBPUT-LL  /$PROJ/website  wa-stats									##==upload wa-stats subdir, and nothing but--!!--
fi
}  2>&1 |tee /tmp/webalize-auto-msgs-$PROJ								##==msgs to stdout, and into /tmp-file for cron-use


exit
==========
== METHOD:
==========
normally we webalize one day's logfile, for preceding day;
however for robustness, we determine the last date that has been webalized, then:
	we do the best we can for all unwebalized days up to curdate-7;
	but stop at any bad day after that, leaving that day and any later days undone;
	eg: on 2009-12-08, newest-logfile:20091207, the 20091201 logfile has just had its last chance, ergo: we stop at bad day among 200912{02,03,04,06,07};
	--WEBGETLOGS-changes on 2010-03-09:  scrapped the "dubious" renaming to "-altfromGDATE",  added need to handle "-hourly";
Notes on logfile names:
	/pix/WEBLOGS/access_log-$DOM.$DATE			<--normal Daily-logfile for $DATE
	/pix/WEBLOGS/access_log-$DOM.$DATE-altfrom$GDATE	<--dubious Daily-logfile for $DATE obtained on $GDATE;  when multiple such we use most recent??
	/pix/WEBLOGS/access_log-$DOM.$DATE-hourly$GTIMESTAMP	<--Hourly-logfile for $DATE obtained at $GTIMESTAMP;  we use catenation of ALL such PLUS normal-Daily!!
To get last date webalized:  we look at newest /pix/WEBSTATS-$PROJ/wa-Monthly_YYYYMM.htm file, last row in "Daily Statistics" table;
note that logfile for $DATE contains records for first 2 hours of following day;

OBSOLETE PARTS:
===============
##==building ARGS, the version that handled dubious-DAILYs, but not HOURLYs:
ARGS=;													##init ARGS-list for webalize-call
for ((DT=WDATEN; DT<DATE; DT=$(dateplusdays $DT +1) ));do						##for each unwebalized day...
  if F=/pix/WEBLOGS/access_log-$DOM.$DT;  [ -e $F ];then ARGS="$ARGS $F"				##add $F to ARGS, for F a normal logfile
  elif ((DT>=$(dateplusdays $DATE -6)            ));then echo "webalize-auto: quit at $DT";  break	##quit at nonexistent/dubious day within last 6
  elif for A in $F-altfrom*;do true;done; [ -e $A ];then ARGS="$ARGS $A"				##add $A to ARGS, for A the newest of the dubious logfiles
  fi
done

CHANGE-LOG:
===========
NOTE: one thing that can go wrong:  when last webalized log has no logrecords from early-AM hours of next day, then it will be fed in twice;
	could happen on ultra low-activity website, or if webhoster-computer down during those hours -- expect it to be very rare, and likely harmless (webalizer kluge);

2009-12-08:  the simplest approach would be to do the webalizing + uploading from webput, so it only happens when I'm ready to upload other changes -- but decided on:
	WEBPUT-LL now supports optional 2nd-param (largely untested);  using that option here to restrict upload to wa-stats dir;
	because automated daily-upload would otherwise be too bloody dangerous wrt unfinished changes;

2009-12-10:  improved MSGS,  webalize just writes to stdout/stderr;  then this script redirects everything from both to tmpfile;
	this solves seeing msgs out-of-order in manual use,  losing chg,etc msgs in cron-use,  also means only ONE /tmp/webalize-XX file per PROJ--!!--
	switched to <tee>;  &>$MSG;cat $MSG  went to  2>&1|tee $MSG  <--nicer for interactive use, unchanged for cron-use (altho a gotcha wrt vars);
==see webalize;  has notes on running weblog-search-strings-report...
==integrating with weblog-search-strings-report:
==	catenate onto monthly-catted-logfile, and run weblog-search-strings-report here, pass param to webalize telling it to do the replacement??
==	OR, that param tells webalize to do such catenating, running, replacement??

2010-03: WEBGETLOGS-changes:  scrapped the "-altfrom" naming of "Dubious" logfiles -- since was doing no good AND HostExcellence has screwed up the naming/numbering...
	however,  WEBGETLOGS is now leaving one or more HOURLYs, named .DATE-hourlyTIMESTAMP, whenever there has been "Restart Problems" at HE;
	when HOURLY(s) exist, they + DAILY are catenated;
	    Need to see examples, but my guess is that lost-data happens when they reboot/Apache-restart that empties the log, ergo don't expect them to overlap==??==