Classify *anything* in a taxonomy

I like Kjartan's proposal a lot. Of course, I proposed a less mature version of this in October: http://list.drupal.org/drupal-devel/2002-October/007901.html

At first glance, it seems that adding a column for Domain is easier than associating URIs directly with terms.

-moshe

----- Original Message -----
From: "Kjartan Mannes"
To:
Sent: Thursday, December 12, 2002 2:11 PM
Subject: Re: [drupal-devel] feature #899 : Classify *anything* in a taxonomy

> Thursday, December 12, 2002, 7:42:59 PM, weitzman @ www . drop . org wrote:
> > Instead of associating only nodes with taxonomy terms, It would be
> > immensely useful to classify URIs instead. This is a superset of our
> > current functionality, also offerring us a way to classify users,
> > comments, feeds, and other non node items.
>
> I'm starting to dislike the current taxonomy API. Just too many support
> functions to do virtually the same thing. Just gonna throw some vague
> ideas out there and see how they land.
>
> Terms should not be linked to nid, but instead a general object. The
> table could look like:
>
> term_links table:
> | context | id | tid
> |----------|--------|----
> | node | 2 | 4
> | url | 2 | 2
> | node | 3 | 5
>
> Then have functions ala:
>
> taxonomy_get_term($conditions) {
> return array of terms matching conditions
> }
>
> taxonomy_get_vocabulary($conditions) {
> return array of vocabs matching conditions
> }
>
> taxonomy_get_synonym($conditions) {
> return array of synonyms matching conditions
> }
>
> taxonomy_get_hierarchy($conditions) {
> return array of hierarchy matching conditions
> }
>
> taxonomy_get_relation($conditions) {
> return array of relations matching conditions
> }
>
> taxonomy_get_link($conditions) {
> // For all links to nodes: $conditions = array("context" => "node");
> // For all links to a specific url: $conditions = array("context" => "url", "id" => 3);
> // For urls linked to a tid: $conditions = array("context" => "url", tid" => 3);
> return array of matching links
> }
>
> taxononmy_set_link($data) {
> // $data = array("context" => "node", "id" => 4, "tid" => 4");
> }
>
> If you added vid to the term_link table you could do pretty much
> anything with those functions. To me it just feels cleaner, and less
> functions to remember. Then again my knowledge of the taxonomy module
> isn't perfect so I don't know if this is a good idea.
>
> --
> Kjartan (http://natrak.net/)

Here is a 7 line patch which enables classifying anything in a vocabulary, not just nodes. This is done by adding a 'domain' column to the term_node table. Usually this column has 'node_nid' as its domain but modules are expected to provide their own domains.

for example, assume a module called feed.module which wants to classify feeds into a taxonomy. you can use 'feed_fid' as your domain and then add feed/term associations at will. The UI for adding these assocations is currently left up to the calling module, and not the taxonomy API. I added a '_taxo()' hook so that these non-node modules may hook into the Vocabulary admin page.

This feature is a start for private taxonomies as well.

In order to use this patch, you must add a column to your term_node table using the following SQL statement

ALTER TABLE term_node ADD domain VARCHAR(255) NOT NULL;

I would think that feeds are a good candidate for being classified as a proof of concept (breyten?) ...

Index: modules/taxonomy.module
===================================================================
RCS file: /cvs/drupal/drupal/modules/taxonomy.module,v
retrieving revision 1.46
diff -u -r1.46 taxonomy.module
--- modules/taxonomy.module	15 Apr 2003 19:50:04 -0000	1.46
+++ modules/taxonomy.module	26 Apr 2003 15:50:20 -0000
@@ -69,7 +69,7 @@
 
 function taxonomy_form_vocabulary($edit = array()) {
   foreach (module_list() as $name) {
-    if (module_hook($name, "node")) {
+    if (module_hook($name, "node") || module_hook($name, "taxo")) {
       $nodetypes[$name] = $name;
     }
   }
@@ -362,8 +362,8 @@
 }
 
 // return array of terms of a node beloging to a particular vocabulary identified by $vid
-function taxonomy_node_get_terms_by_vocabulary($nid, $vid, $key = "tid") {
-  $result = db_query("SELECT t.* FROM term_data t, term_node r WHERE t.tid = r.tid AND t.vid = '%d' AND r.nid = '%d' ORDER BY weight", $vid, $nid);
+function taxonomy_node_get_terms_by_vocabulary($nid, $vid, $key = "tid", $domain = "node_nid") {
+  $result = db_query("SELECT t.* FROM term_data t, term_node r WHERE t.tid = r.tid AND t.vid = '%d' AND r.nid = '%d' AND r.domain = '%s' ORDER BY weight", $vid, $nid, $domain);
   $terms = array();
   while ($term = db_fetch_object($result)) {
     $terms[$term->$key] = $term;
@@ -372,11 +372,11 @@
 }
 
 // return array of terms of a node
-function taxonomy_node_get_terms($nid, $key = "tid") {
+function taxonomy_node_get_terms($nid, $key = "tid", $domain = "node_nid") {
   static $terms;
 
   if (!isset($terms[$nid])) {
-    $result = db_query("SELECT t.* FROM term_data t, term_node r WHERE r.tid = t.tid AND r.nid = '%d' ORDER BY weight, name", $nid);
+    $result = db_query("SELECT t.* FROM term_data t, term_node r WHERE r.tid = t.tid AND r.nid = '%d' AND r.domain = '%s' ORDER BY weight, name", $nid, $domain);
     $terms[$nid] = array();
     while ($term = db_fetch_object($result)) {
       $terms[$nid][$term->$key] = $term;
@@ -386,12 +386,13 @@
 }
 
 // save terms of a node
-function taxonomy_node_save($nid, $terms) {
+function taxonomy_node_save($nid, $terms, $domain = "node_nid") {
+  
   taxonomy_node_delete($nid);
 
   if ($terms) {
     foreach ($terms as $term) {
-      db_query("INSERT INTO term_node (nid, tid) VALUES ('%d', '%d')", $nid, $term);
+      db_query("INSERT INTO term_node (nid, tid, domain) VALUES ('%d', '%d', '%s')", $nid, $term, $domain);
     }
   }
 }
 }
 
 // return array of terms of a node
-function taxonomy_node_get_terms($nid, $key = "tid") {
+function taxonomy_node_get_terms($nid, $key = "tid", $domain = "node_nid") {
   static $terms;
 
   if (!isset($terms[$nid])) {
-    $result = db_query("SELECT t.* FROM term_data t, term_node r WHERE r.tid = t.tid AND r.nid = '%d' ORDER BY weight, name", $nid);
+    $result = db_query("SELECT t.* FROM term_data t, term_node r WHERE r.tid = t.tid AND r.nid = '%d' AND r.domain = '%s' ORDER BY weight, name", $nid, $domain);
     $terms[$nid] = array();
     while ($term = db_fetch_object($result)) {
       $terms[$nid][$term->$key] = $term;
@@ -386,12 +386,13 @@
 }
 
 // save terms of a node
-function taxonomy_node_save($nid, $terms) {
+function taxonomy_node_save($nid, $terms, $domain = "node_nid") {
+  
   taxonomy_node_delete($nid);
 
   if ($terms) {
     foreach ($terms as $term) {
-      db_query("INSERT INTO term_node (nid, tid) VALUES ('%d', '%d')", $nid, $term);
+      db_query("INSERT INTO term_node (nid, tid, domain) VALUES ('%d', '%d', '%s')", $nid, $term, $domain);
     }
   }
 }

Breyten has given positive feedback for the patch here. Anyone else interested in this? It is ready for CVS consideration IMO.

In order to view non-nodes which have been classified in the taxonomy, you have to roll your own view module which implements the txonomy_node_get_terms() function. That function name is misleading because after this patch, it can operate on non-nodes as well - see the 3rd parameter. But changing the function name means changing many other modules so I choose to keep it for now.

You cannot simply use the taxonomy/view/or/x syntax because taxonomy doesn't know how to render your 'anything' items.

And I should add that you have to roll your own UI for associating your 'anything' with taxonomy terms. This patch adjusts the core API, not user interface.

This discussion about how classification should work is nice, but we have to talk about code at some point. Specifically, we have talked for more than year about storing comments as nodes. But no patch for this has ever emerged. Maybe it is too hard, or maybe noone is motivated enough. Comment.module just underwent a significant overhaul, and it didn't include transition to node storage.

My point is that we have, in this project item, a working patch. In all other scenarios, we have hot air. This difference is very important. At some point, saying 'everything should be a node' is much less helpful than producing code code which gets us there.

I'n not trying to belittle an important debate, I just want to emphasize the importance of working code over theoretical discussion.

anyone interested in reviving this effort? i think its still worth exploring.

One possible way forward is for taxonomy to become a 'field' in #265604: Fields in core sense. Thats a ways into the future though.