Iñaki Baz Castillo writes:

The problem is that it would require perform N regular expression matching and such expressions will be very long (all the country prefixes in the world). The main problem is that some groups would contain prefixes starting with, for example, 12, while in other group there is the prefix 1 (an the very same in multiple cases, so doing it manually would be too much manual work and any change in the world would require hard modifications).

perhaps you could use dialplan module. it there is overlapping regular expressions, one is selected based on its priority.

-- juha

Iñaki Baz Castillo writes:

Thanks. I don't need different priorities, I just need that, in case of multiple matching (i.e. ^1 , ^10) the longest matching wins. Does the dialplan module allow it?

try by giving rule ^1 lower priority than ^10.

-- juha

Hello,

On 6/6/11 10:39 AM, Iñaki Baz Castillo wrote:

2011/6/6 Daniel-Constantin Mierlamiconda@gmail.com:

...

one option that comes in my mind is to break it in several if conditions, matching one group at a time, written in the order of preferences.

The problem is that it would require perform N regular expression matching and such expressions will be very long (all the country prefixes in the world). The main problem is that some groups would contain prefixes starting with, for example, 12, while in other group there is the prefix 1 (an the very same in multiple cases, so doing it manually would be too much manual work and any change in the world would require hard modifications).

I'm evaluating building a custom function for this purpose.

maybe dialplan module could give you a solution -- you can play with priorities for matching rules.

Cheers, Daniel

2011/6/6 Henning Westerholt henning.westerholt@1und1.de:

if you just want to have the longest match on a prefix tree, this is what carrierroute was build for.

...

In case $rU = "999999" I need to get "null", in case $rU = "1000022323" I need to get "1".

Its works a bit differently, but this is similar to the logic that we use. You've a bunch of specific prefixes with special routing, normal country prefixes that you want to route differently and then a default routing for the rest

Thanks Henning. I do know that CR and LCR operate with longest prefix matching, but I don't want this feature now for routing purposes, but for blacklist purposes. So using dialplan module with higher priorities in longer prefixes would be valid in my case.

Thanks a lot.

2011/6/6 Henning Westerholt henning.westerholt@1und1.de:

We use the userblacklist module for this, it works with the same longest prefix match logic as cr.

Ops, I missed that module :) I'll take a look to it. Thanks.

On Monday 06 June 2011, Iñaki Baz Castillo wrote:

It's not valid for my custom case as I need a function returning an id of the matched groups. This is, I have various geo-zone groups:

1: Europe (prefixes of Europe) 2: North America (prefixes of North America) ...

Each client has a boolean field to set permission for calls to each geo-zones.

So I think it's better for me to use dialplan module and return the geo-zone id in the "attrs" AVP.

Yes - the dialplan is much more flexible of course. :-)

Henning

2011/6/6 Juha Heinanen jh@tutpro.com:

do both of them also support regular expressions, what inaki was asking for?

Dialplan module allows setting a regular expression per table entry (by setting the column match_op to 1 (regexp).

2011/6/6 Juha Heinanen jh@tutpro.com:

sure, but i was asking about cr and the other module that henning was recommending.

Well, for sure cr and userblacklist implements regular expression matching by taking the prefix from the DB and internally using "^PREFIX" (I expect). Same as LCR.

The fact is that I don't need more regular expressions than that. This is, I don't need to set a prefix like "(1|2[0123]{2})". :)

2011/6/6 David Villasmil david.villasmil.work@gmail.com:

I connected to mysql and ran something like: select * from routes where '$rU' like concat(areacode,'%') order by len(areacode) desc limit 1;

Hi David, that would work indeed, but I prefer not to saturate the DB with such a query which cannot use table indexes. My kamailio handles 2000-3000 concurrent calls and such calls come from callcenters so they are very "aggressive" (maybe 200-300 calls in the same second), so querying the database with a low performance query is not an option in my case.

Thanks.

2011/6/6 David Villasmil david.villasmil.work@gmail.com:

I had much the same traffic, and with a good index that wouldn't be a problem, IMHO.

Hi David, your query cannot make usage of table indexes:

select * from routes where '$rU' like concat(areacode,'%') order by len(areacode) desc limit 1;

You are using "LIKE" and concat(TABLE_COLUMN...) so you are forcing the database server to read *all* the values of "areacode" (all the rows), create a new string by concatening "%" at the end, and then performing a regular expressión (LIKE "XXX%") for *every* resulting values. This can never use an index.

You can check it by yourself by entering into mysql console:

DESCRIBE select * from routes where '999999999' like concat(areacode,'%') order by len(areacode) desc limit 1;

You will realize that no index is being used.

You could also load the table in memory, can't be faster than that.

Yes, but for that a custom module is required :)

Of course, doing it in the config script is faster, but it limits your flexibility... just a thought...

Humm, in fact not. I can manage the table content via a web interface (or whatever) and running a MI "reload" command for the used module so the table content is read again into memory. This is the same as I do with LCR module and others.

Regards.

2011/6/6 Daniel-Constantine Mierla miconda@gmail.com:

Btw, very lightweight module for longest prefix matching is mtree. It has option to return a value in a cfg variable, that can be the group id for example. It does caching in memory trees, so it is really fast in matching.

I also missed that module. Let's take a look to it.

Thanks.

Hello,

On 6/6/11 3:29 PM, Iñaki Baz Castillo wrote:

2011/6/6 Iñaki Baz Castilloibc@aliax.net:

...

2011/6/6 Daniel-Constantine Mierlamiconda@gmail.com:

...

Btw, very lightweight module for longest prefix matching is mtree. It has option to return a value in a cfg variable, that can be the group id for example. It does caching in memory trees, so it is really fast in matching.

I also missed that module. Let's take a look to it.

Hi Daniel, sounds really specific for which I need :) Let me ask some doubts:

• You say "It has option to return a value in a cfg variable, that can

be the group id for example". Do you mean pv_value AVP?: http://kamailio.org/docs/modules/3.1.x/modules/mtree.html#id2845134

yes, but it can be any kind of writable variable, not restricted to AVP -- see the example in the readme where it is a $var()

Would that AVP be filled with the content of the "value" column for the matched row?

Exactly

• Which is the meaning of "payload" here?: http://kamailio.org/docs/modules/3.1.x/modules/mtree.html#id2807543

when it is 0, will return the value as string in varibale. The goal is to parse in cache the value into some internal structure. Right now there is also the option of 1, expecting some 'id=value', both being integers, with the purpose of returning all matches for the prefix ordered by value. It hasn't be tested extensively and thus not documented - initially work was on hold to get first xavps so the structure can be returned nicely to config, then lack of time kept it posponed.

• What does mt_match() return in case there is match? and when there is not?

True and false.

Cheers, Daniel

Thanks a lot.

2011/6/6 Iñaki Baz Castillo ibc@aliax.net:

Thanks Daniel, please let me one question more: does the mtree module allow setting two different tables (with maybe also different columns names)? The function mt_match(mtree, pv, mode) seems to allow, as first argument, the "model/table" to use, but I see no way of defining different models, is it?

Ok, it seems that using "mtree" parameter does the job (but columns must have same name in all the tables):

modparam("mtree", "mtree", "name=mytable;dbtable=routes;type=0;")

On 6/6/11 5:26 PM, Iñaki Baz Castillo wrote:

2011/6/6 Iñaki Baz Castilloibc@aliax.net:

...

...

Thanks Daniel, please let me one question more: does the mtree module allow setting two different tables (with maybe also different columns names)? The function mt_match(mtree, pv, mode) seems to allow, as first argument, the "model/table" to use, but I see no way of defining different models, is it?

Ok, it seems that using "mtree" parameter does the job (but columns must have same name in all the tables):

modparam("mtree", "mtree", "name=mytable;dbtable=routes;type=0;")

Sorry for so many questions but, what is "tname_column"?:

3.4. tname_column (string) Name of 'tname' column.

The table scheme clearly shows id, tprefix and tvalue columns, tprefix_column and tvalue_column parameters can be used, as usual, to rename such columns but, what is "tname_column"?

The module can load on db table in a memory tree or one db table in many trees. For the second case, the db table name is a module parameter and in addition it must have an extra column to specify the memory tree name.

Cheers, Daniel

On 6/6/11 5:23 PM, Iñaki Baz Castillo wrote:

2011/6/6 Iñaki Baz Castilloibc@aliax.net:

...

Thanks Daniel, please let me one question more: does the mtree module allow setting two different tables (with maybe also different columns names)? The function mt_match(mtree, pv, mode) seems to allow, as first argument, the "model/table" to use, but I see no way of defining different models, is it?

Ok, it seems that using "mtree" parameter does the job (but columns must have same name in all the tables):

modparam("mtree", "mtree", "name=mytable;dbtable=routes;type=0;")

yes columns names are configurable per module -- maybe in the future will be per mtree.

Meanwhile maybe you can use mysql views to get to the column names expected by mtree.

Cheers, Daniel