On 18-Feb-02 Ted Harding wrote:

Hi Jenny,

3. I don't know how one "registers" an access to a web page

within Linux. Suppose, however, that the access updates some file called "acces_log". Then the script in (1) could be

touch watch_access_log while true ; do if [ access_log -nt watch_access_log ] ; then cat appended_file >> other_file fi sleep 10 touch watch_access_log done

It occurs to me that, while you can set the "sleep" to anything you like down to 1 second (but you can't set it smaller), you might have two or more web-page accesses during the sleep period; and you wanted to do the append for every access.

The only way I can see how to get round that would be for the web-access to trigger a signal to another program which then immediately did the append. I don't know how you would arrange for the signal to be sent. However, you can start up a shell which can detect signals by using the "trap" command in bash, on the lines of the (outline) script

#!/bin/bash trap "command_to_append_file" SIGINT while true; do <nothing_much> ; done

For example, try the following little script (call it "temp"):

#!/bin/bash trap "ls -l" SIGINT while true ; do true > /dev/null ; done

and then

chmod 755 temp

1. Make sure you have another window from which you can do

killall -15 temp

Then start temp with the command

temp

Now keep trying to interrupt it with ^C -- each time, you will get the output of "ls -l". When you're happy about that, kill it as above from the other window.

To make it work in the background, do

temp &

and now, from anywhere, you can

killall -2 temp

which will send it the SIGINT signal, resulting in a prompt "ls -l"; and this will continue until you "killall -15 temp".

You can use this mechanism instead of the "watch" loop in my other mail provided you have a way of getting the web-page access to send the signal to the above program. That I can't help with.

The only thing you need to change in the script is to substitute "cat appended_file >> other_file" for "true > /dev.null".

Alternatively, if you reckon that "sleep 1" (or whatever) is safe from multiple wep-page accesses, then you can stick to the other script which at least is independent of getting something else to send the signal, since all it does is watch a file.

Ted.

-------------------------------------------------------------------- E-Mail: (Ted Harding) Ted.Harding@nessie.mcc.ac.uk Fax-to-email: +44 (0)870 167 1972 Date: 18-Feb-02 Time: 14:26:55 ------------------------------ XFMail ------------------------------

On 18-Feb-02 Eric Sellin wrote:

Much easier to do a "tail -f" on the Web server's log file, read lines in a loop and do whatever we want to do for each line.

Assuming that's how you tell a page has been accessed, then this is a neat way of doing it.

BTW, I think it's time Jenny told in a bit more detail what she needs to do when the web-page is accessed. Is it any web-page, with a different action for each web page? Or is it the same action for any access? Is it simply access to the web-site as a whole?

In any case, following up on Eric's suggestion above, 'awk' would be a very good way of doing it, since it is designed to be a program which automatically responds line-by-line to line-by-line input. This would go on the following lines:

tail -f -l 1 access_log | awk 'BEGIN{next} {system("cat append_file >> other-file")}'

'tail' tracks access_log, outputting the most recent line added to it. The BEGIN action discards the stale initial line which would be read from access_log on startup. Thereafter, each new line output by tail will trigger the append.

However, this assumes that a web-page access appends one line only to access_log. and would fail if a web-page access generated several lines in access_log, since then each of these would trigger the append. In that case one would need to make the 'awk' command a bit more intelligent.

For instance, suppose -- no matter how many lines a single web-page access generated -- there was amongst these a single line containing a certain pattern. Then you can make 'awk' respond only to lines containing this pattern, as in

tail -f -l 1 access_log | awk 'BEGIN{next} /pattern/{system("cat append_file >> other-file")}'

Since (in the 'awk' command) /pattern/ can be a regular expression, this also allows several different (but equivalent) forms of pattern to be recognised as indicating that a web-page has been accessed; and, should a different action be required for some, this can also be accomodated; as in:

tail -f -l 1 access_log | awk 'BEGIN{next} /pattern1/{system("action 1")} /pattern2/{system("action 2")} .... /patternN/{system("action N")}'

And so it goes ... Ted.

-------------------------------------------------------------------- E-Mail: (Ted Harding) Ted.Harding@nessie.mcc.ac.uk Fax-to-email: +44 (0)870 167 1972 Date: 18-Feb-02 Time: 15:46:48 ------------------------------ XFMail ------------------------------