[patch] Flexinode + Taxonomy support in node_import

As discussed in http://drupal.org/node/18207 I have prepared a patch to include support for taxonomy and flexinode in node_import.

List support for all columns

If you add "[]" to the end of a fieldname (first row), the module will convert the data in this column to an array of values. Example:

"title","name","type","list[]"
"List test","admin","page","1|2|3"

results in:

stdClass Object
(
    [title] => List test
    [name] => admin
    [type] => page
    [body] => 
    [list] => Array
        (
            [0] => 1
            [1] => 2
            [2] => 3
        )
    ...
)

You can override the used seperator (defaults to a vertical bar "|"), but inserting it inside the square brackets. For example, to use a semicolon (";") instead of vertical bar ("|"), you can do:

"title","name","type","list[;]"
"List test","admin","page","1;2;3"

I'm not sure if this is a feature of "CSV" or just of "fgetcsv", but you can specify a newline as seperator too:

"title","name","type","list[
]"
"List test","admin","page","1
2
3"

These last two examples will create the same node as the first example does.

Special handling of taxonomy field

The initial reason for list-support was to import taxonomy terms. For example:

"title","name","type","taxonomy[]"
"Taxonomy test","admin","page","1|2|3"

Looking up the numbers ("tid") for each term is probably error-prune, so the module handles a column with name "taxonomy" specially: if the value of the field is not numeric, we check if the string is not a term defined in a vocabulary. If it is, the value is replaced by the "tid" of this term. For example:

"title","name","type","taxonomy[]"
"Taxonomy test","admin","page","An example term|Another term"

results in:

stdClass Object
(
    [title] => Taxonomy test
    [name] => admin
    [type] => page
    [body] => 
    [taxonomy] => Array
        (
            [0] => 5
            [1] => 6
        )
    ...
)

BTW: if a term is not found, the string is kept as is. The taxonomy module ignores it. For example:

"title","name","type","taxonomy[]"
"Taxonomy test","admin","page","An example term|Another term|No such term"

becomes:

stdClass Object
(
    [title] => Taxonomy test
    [name] => admin
    [type] => page
    [body] => 
    [taxonomy] => Array
        (
            [0] => 5
            [1] => 6
            [2] => No such term
        )
    ...
)

And you can mix numbers with strings too:

"title","name","type","taxonomy[]"
"Taxonomy test","admin","page","An example term|1|No such term"

creates:

stdClass Object
(
    [title] => Taxonomy test
    [name] => admin
    [type] => page
    [body] => 
    [taxonomy] => Array
        (
            [0] => 5
            [1] => 1
            [2] => No such term
        )
    ...
)

Special handling of flexinode types

With the node_import module you could import flexinodes, you just had to know how. For example, if we define a flexinode ("flexinode-1" = "link") with one field ("flexinode_1" = "url"), we need to create a CSV file as:

"title","name","type","ctype_id","flexinode_1"
"Flexinode test","admin","flexinode-1","1","http://example.org"

I find it easier to write something like:

"title","name","type","url"
"Flexinode test","admin","link","http://example.org"

This patch enable this. The node that gets imported is:

stdClass Object
(
    [title] => Flexinode test
    [name] => admin
    [type] => flexinode-1
    [body] => 
url:
http://example.org
    [url] => http://example.org
    [ctype_id] => 1
    [flexinode_1] => http://example.org
    ...
)

Observe that:

• "$node->type" is set to "flexinode-1"
• "$node->ctype_id" is set to the correct value automagically
• both the original "url" and the new "flexinode_1" are present

BTW: the field "flexinode_1" is only added to "$node" when the "type" is a flexinode type that has a field with this label. So when "type" is not a flexinode type or when "type" is a flexinode type without an "url" field, then no "flexinode_1" field is added.

The patch is probably not perfect. If there is something wrong with it (either bug or "Just Plain Wrong(tm)") I like to hear about it.

Some things that might be added (maybe by me):

• if the term is not found, issue a warning?
• if the term is not found, create it?
• if "taxonomy_get_term_by_name" returns more than one term... maybe set them all? maybe issue a warning?
• if the node type is not found, issue a warning?
• cache the lookup of valid fields, so no lookup of "field_ids" is done for each row, but only once for each flexinode-type.

moshe, can you let me know if this is likely to be applied? or what is needed so it can be applied?

Kind regards,
Robrecht