<?xml version="1.0" encoding="latin1" ?>
<!DOCTYPE erlref SYSTEM "erlref.dtd">

<erlref>
  <header>
    <copyright>
      <year>1997</year><year>2009</year>
      <holder>Ericsson AB. All Rights Reserved.</holder>
    </copyright>
    <legalnotice>
      The contents of this file are subject to the Erlang Public License,
      Version 1.1, (the "License"); you may not use this file except in
      compliance with the License. You should have received a copy of the
      Erlang Public License along with this software. If not, it can be
      retrieved online at http://www.erlang.org/.
    
      Software distributed under the License is distributed on an "AS IS"
      basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
      the License for the specific language governing rights and limitations
      under the License.
    
    </legalnotice>

    <title>snmp_index</title>
    <prepared></prepared>
    <responsible></responsible>
    <docno></docno>
    <approved></approved>
    <checked></checked>
    <date></date>
    <rev></rev>
    <file>snmp_index.xml</file>
  </header>
  <module>snmp_index</module>
  <modulesummary>Abstract Data Type for SNMP Indexing</modulesummary>
  <description>
    <p>The module <c>snmp_index</c> implements an Abstract
      Data Type (ADT) for an SNMP
      index structure for SNMP tables.  It is implemented as an ets
      table of the ordered_set data-type, which means that all operations are
      O(log n).  In the table, the key is an ASN.1 OBJECT
      IDENTIFIER.
      </p>
    <p>This index is used to separate the implementation of the SNMP
      ordering from the actual implementation of the table.  The SNMP
      ordering, that is implementation of GET NEXT, is implemented in this
      module.
      </p>
    <p>For example, suppose there is an SNMP table, which is best
      implemented in Erlang as one process per SNMP table row.  Suppose
      further that the INDEX in the SNMP table is an OCTET STRING.  The
      index structure would be created as follows:
      </p>
    <code type="none">
snmp_index:new(string)
    </code>
    <p>For each new process we create, we insert an item in an
      <c>snmp_index</c> structure:
      </p>
    <code type="none"><![CDATA[
new_process(Name, SnmpIndex) ->
  Pid = start_process(),
  NewSnmpIndex = 
    snmp_index:insert(SnmpIndex, Name, Pid),
  <...>
    ]]></code>
    <p>With this structure, we can now map an OBJECT IDENTIFIER in
      e.g. a GET NEXT request, to the correct process:
      </p>
    <code type="none">
get_next_pid(Oid, SnmpIndex) ->
  {ok, {_, Pid}} = snmp_index:get_next(SnmpIndex, Oid),
  Pid.
    </code>
  </description>

  <section>
    <title>Common data types</title>
    <p>The following data types are used in the functions below:
      </p>
    <list type="bulleted">
      <item>
        <p><c>index()</c></p>
      </item>
      <item>
        <p><c>oid() = [byte()]</c></p>
      </item>
      <item>
        <p><c>key_types = type_spec()  |  {type_spec(), type_spec(), ...}</c></p>
      </item>
      <item>
        <p><c>type_spec() = fix_string | string | integer</c></p>
      </item>
      <item>
        <p><c>key() = key_spec() | {key_spec(), key_spec(), ...}</c></p>
      </item>
      <item>
        <p><c>key_spec() = string() | integer()</c></p>
      </item>
    </list>
    <p>The <c>index()</c> type denotes an snmp index structure.
      </p>
    <p>The <c>oid()</c> type is used to represent an ASN.1 OBJECT
      IDENTIFIER.
      </p>
    <p>The <c>key_types()</c> type is used when creating the
      index structure, and the <c>key()</c> type is used when inserting
      and deleting items from the structure.
      </p>
    <p>The <c>key_types()</c> type defines the types of the SNMP INDEX
      columns for the table.  If the table has one single INDEX column,
      this type should be a single atom, but if the table has multiple
      INDEX columns, it should be a tuple with atoms.
      </p>
    <p>If the INDEX column is of type INTEGER, or derived from
      INTEGER, the corresponding type should be <c>integer</c>.  If it
      is a variable length type (e.g. OBJECT IDENTIFIER, OCTET STRING),
      the corresponding type should be <c>string</c>.  Finally, if the
      type is of variable length, but with a fixed size restriction
      (e.g. IpAddress), the corresponding type should be
      <c>fix_string</c>.
      </p>
    <p>For example, if the SNMP table has two INDEX columns, the first
      one an OCTET STRING with size 2, and the second one an OBJECT
      IDENTIFER, the corresponding <c>key_types</c> parameter would be
      <c>{fix_string, string}</c>.
      </p>
    <p>The <c>key()</c> type correlates to the <c>key_types()</c>
      type. If the <c>key_types()</c> is a single atom, the
      corresponding <c>key()</c> is a single type as well, but if the
      <c>key_types()</c> is a tuple, <c>key</c> must be a tuple of the
      same size.
      </p>
    <p>In the example above, valid <c>keys</c> could be <c>{"hi", "mom"}</c> and <c>{"no", "thanks"}</c>, whereas <c>"hi"</c>,
      <c>{"hi", 42}</c> and <c>{"hello", "there"}</c> would be invalid.</p>
    <warning>
      <marker id="1"></marker>
      <p>All API functions that update the index return a <c>NewIndex</c>
        term. This is for backward compatibility with a previous
        implementation that used a B+ tree written purely in Erlang for 
        the index. The <c>NewIndex</c> return value can now be ignored. 
        The return value is now the unchanged table identifier for the 
        ets table.</p>
      <p>The implementation using ets tables introduces a semantic
        incompatibility with older implementations. In those older 
        implementations, using pure Erlang terms, the index was garbage 
        collected like any other Erlang term and did not have to be 
        deleted when discarded. An ets table is deleted only when the 
        process creating it explicitly deletes it or when the creating 
        process terminates.</p>
      <p>A new interface <c>delete/1</c> is now added to
        handle the case when a process wants to discard an index table
        (i.e. to build a completely new). Any application using
        transient snmp indexes has to be modified to handle this.</p>
      <p>As an snmp adaption usually keeps the index for the whole of the
        systems lifetime, this is rarely a problem.</p>
    </warning>
  </section>
  <funcs>
    <func>
      <name>delete(Index) -> true</name>
      <fsummary>Delete an index table</fsummary>
      <type>
        <v>Index = NewIndex = index()</v>
        <v>Key = key()</v>
      </type>
      <desc>
        <p>Deletes a complete index structure (i.e. the ets table
          holding the index). The index can no longer be referenced
          after this call. See the <seealso marker="#1">warning note</seealso>
          above.</p>
      </desc>
    </func>
    <func>
      <name>delete(Index, Key) -> NewIndex</name>
      <fsummary>Delete an item from the index</fsummary>
      <type>
        <v>Index = NewIndex = index()</v>
        <v>Key = key()</v>
      </type>
      <desc>
        <p>Deletes a key and its value from the index structure.
          Returns a new structure.</p>
      </desc>
    </func>
    <func>
      <name>get(Index, KeyOid) -> {ok, {KeyOid, Value}} | undefined</name>
      <fsummary>Get the item with <c>KeyOid</c></fsummary>
      <type>
        <v>Index = index()</v>
        <v>KeyOid = oid()</v>
        <v>Value = term()</v>
      </type>
      <desc>
        <p>Gets the item with key <c>KeyOid</c>.  Could be used from
          within an SNMP instrumentation function.</p>
      </desc>
    </func>
    <func>
      <name>get_last(Index) -> {ok, {KeyOid, Value}} | undefined</name>
      <fsummary>Get the last item in the index structure</fsummary>
      <type>
        <v>Index = index()</v>
        <v>KeyOid = oid()</v>
        <v>Value = term()</v>
      </type>
      <desc>
        <p>Gets the last item in the index structure.</p>
      </desc>
    </func>
    <func>
      <name>get_next(Index, KeyOid) -> {ok, {NextKeyOid, Value}} | undefined</name>
      <fsummary>Get the next item</fsummary>
      <type>
        <v>Index = index()</v>
        <v>KeyOid = NextKeyOid = oid()</v>
        <v>Value = term()</v>
      </type>
      <desc>
        <p>Gets the next item in the SNMP lexicographic ordering,
          after <c>KeyOid</c> in the index structure.  <c>KeyOid</c>
          does not have to refer to an existing item in the index.</p>
      </desc>
    </func>
    <func>
      <name>insert(Index, Key, Value) -> NewIndex</name>
      <fsummary>Insert an item into the index</fsummary>
      <type>
        <v>Index = NewIndex = index()</v>
        <v>Key = key()</v>
        <v>Value = term()</v>
      </type>
      <desc>
        <p>Inserts a new key value tuple into the index structure.  If
          an item with the same key already exists, the new <c>Value</c>
          overwrites the old value.</p>
      </desc>
    </func>
    <func>
      <name>key_to_oid(Index, Key) -> KeyOid</name>
      <fsummary>Convert a key to an OBJECT IDENTIFIER</fsummary>
      <type>
        <v>Index = index()</v>
        <v>Key = key()</v>
        <v>KeyOid = NextKeyOid = oid()</v>
      </type>
      <desc>
        <p>Converts <c>Key</c> to an OBJECT IDENTIFIER.</p>
      </desc>
    </func>
    <func>
      <name>new(KeyTypes) -> Index</name>
      <fsummary>Create a new snmp index structure</fsummary>
      <type>
        <v>KeyTypes = key_types()</v>
        <v>Index = index()</v>
      </type>
      <desc>
        <p>Creates a new snmp index structure.  The <c>key_types()</c>
          type is described above.</p>
      </desc>
    </func>
  </funcs>
  
</erlref>