FileFilters not being invoked?

Addition: I'm also chmoding the temporary conf file to 0777 after it is created on-the-fly, plus the Swish-e output suggests that my problem does not lie with the config file.

I can confirm Kenny_au's estimation of the situation. It can be run from bash w/o the dbl quotes and catdoc works. It (swish engine, not drupal mod) also gives the following error if I remove the double quotes:

Indexing abroted. err: FileFilter: requires "extension" "filter" "[options]"

Yes, the spelling error was funny, but that was about four thousand errors ago. So, how can you pass the "[options]" arg without double quotes? Well, I removed them, to no avail. I added braces instead, to no avail (yeah, this was in desperation). How else can I pass a string with spaces? I ended up passing the '%p' arg and catdoc worked fine. Let me restate catdoc worked fine!. I have yet to try the other filters.

As for the link's filepath, I think that's just something that needs to be correctly parsed from swish's results. In other words, I think it's the module's fault. I'll see if I can't get this to work correctly. If I do post a patch, it'll be messy, cuz I've hacked this module from dusk till dawn.

Wow, swish is returning almost nothin' after that search command is executed. I truly thought filepath would be passed back... perhaps we just need to ask for it? Unfortunately, I don't know the engine very well (yet).

I think another huge task to be done with this module is access rights checking, unless I missed a whole pantload of code somewhere? If I can get the resources to write in some additional functionality, I'll certainly post.

I am fool. Of course swish is sending back the right path. The module is killing it. I'll work on this a bit later and post.

Here's a couple of lines of code that seem to cure the filepath problem. This is not tested, so beware. Insert these lines in function swish_search() right before the line $basename .= basename(substr($result, $k+1, $i-$k-1));

$basename = substr($result, $k+1,$i-$k-1);
$basename = str_replace(variable_get('file_directory_path', 'files'),'',$basename);
$basename = str_replace(basename(substr($result, $k+1, $i-$k-1)),'',$basename);

Aw shoot. One last thing, change the line:

$basename = basename(substr($result, $k+1, $i-$k-1));

to be:

$basename .= basename(substr($result, $k+1, $i-$k-1));

Kenny_au, are you using public downloads?

As for catdoc, I'm using your code from #5!! Or are you looking for the changes I made to the module to get there?

Again, I'll probably be in and out of this thread, but eventually I'll fix this. If you have specific deadlines, let me know. I might get a chance this weekend or next to commit some of these hacks into real code. Reason being, my job is completely out of hand. I have that boss--you know, the one whom you show your new module that runs 543 lines of regex to add 2% relevance to search results, and he replies, "Isn't that logo a couple of pixels too far to the left?"

I couldn't let this one rest. Here's some new code to get the $basename variable. I don't have my stuff set to public downloads, so I really wasn't able to test this at all. I just parsed the $basename to not have the filepath nor the leading slash. If it fails, just post your search output link, like you did last time.

So, scratch the other lines of code and go with this (so these are the only lines that use $basename, aside from the final building of the link in the line that starts $link)

$basename = substr($result, $k+1,$i-$k-1);
$basename = str_replace(variable_get('file_directory_path', 'files'),'',$basename);
$basename = trim(substr($basename, 1, strlen($basename)-1));

That should yield: 1) a filename with 2) any filepath prefix that may exist (not below public_html) and 3) no leading slash

As for your other questions, I think the google teaser idea is fantastic. If you can figure out the command to send swish-e to return some sort of teaser, I'll take care of the balance--though highlighting the search word might get hairy, or might be a bunch of code, we'll see. I'm not sure swish-e can do this, but it seems pretty robust, so it's probably just that we're not asking it to do so.

My take on your version 1.3 comment is this: At the time it came out, v1.3 was probably working swimmingly. It's just that either the engine or drupal evolved faster than the module did, thus breaking the conf files. Actually, there are even more moving pieces than I'm addressing, given that swish-e is external to drupal. Incidentally, this is totally speculation.

Let me know how your links turn out on the search page.

Interestingly enough, the reason my parsing isn't doing what it is supposed to do is because I hacked the mod elsewhere. I'll post as soon as I can write something that isn't completely broken.

So, it took me about three hours to un-hack all the work I'd done in the original troubleshooting of the module. Then, it took me about five minutes to change the string parsing. Phew (wiping sweat from brow).

I don't have uploads going to more than one folder, so all I can verify is that the outputted link does have the extra node(s) in the filepath when I add child folders under my upload folder (and index some files in that child folder). I did test w/public and private, and it seemed to work, but I sort of feel blind since I only have one upload folder. What module are you using for uploads?

Let me know how your results work. Module attached.

Ignore that last upload, as it had some debug code in it. This should work better.

What! Really?!

As of late, I've really juiced this thing up in my install. Of most interest: uses db_rewrite logic to dictate node access within the search (incredibly for my needs), but also becomes dependent on the core upload module in doing so. Also, I snagged some great icons from the silk collection (open) and use the same theme function as the regular search, so as to keep a consistent look/feel. If any of the above interest you, contact me over email and I'll zip and send your way.

Cheers!

I almost forgot to mention: the last module I uploaded needs the tweaks that you suggested for filefilters for pdftotext and unrtf.