File: tablesDatabases.html

package info (click to toggle)
erlang-doc-html 1%3A11.b.2-1
links: PTS
area: main
in suites: etch, etch-m68k
size: 23,284 kB
ctags: 10,724
sloc: erlang: 505; ansic: 323; makefile: 62; perl: 61; sh: 45
file content (536 lines) | stat: -rw-r--r-- 18,262 bytes
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<!-- This document was generated using DocBuilder 3.3.3 -->
<HTML>
<HEAD>
  <TITLE>Tables and databases</TITLE>
  <SCRIPT type="text/javascript" src="../../doc/erlresolvelinks.js">
</SCRIPT>
</HEAD>
<BODY BGCOLOR="#FFFFFF" TEXT="#000000" LINK="#0000FF" VLINK="#FF00FF"
      ALINK="#FF0000">
<CENTER>
<A HREF="http://www.erlang.se"><IMG BORDER=0 ALT="[Ericsson AB]" SRC="min_head.gif"></A>
</CENTER>
<A NAME="4"><!-- Empty --></A>
<H2>4 Tables and databases</H2>
<A NAME="4.1"><!-- Empty --></A>
<H3>4.1 Ets, Dets and Mnesia</H3>

<P> All examples using Ets has an corresponding example in
Mnesia. In general all Ets examples also applies to Dets tabels.<A NAME="4.1.1"><!-- Empty --></A>
<H4>4.1.1 Select/Match operations</H4>

<P>Select/Match operations on Ets and Mnesia tables can become
        very expensive operations. They will have to scan the whole
        table that may be very large. You should try to structure your
        data so that you minimize the need for select/match
        operations. However if you need a select/match operation this
        will be more efficient than traversing the whole table using
        other means such as <CODE>tab2list</CODE> and <CODE>mnemosyne</CODE>. Examples of this
        and also of ways to avoid select/match will be provided in
        some of the following sections. From R8 the functions
        <CODE>ets:select/2</CODE> and <CODE>mnesia:select/3</CODE> should be preferred over
        <CODE>ets:match/2</CODE>,<CODE>ets:match_object/2</CODE> and <CODE>mnesia:match_object/3</CODE>.

<P>
<TABLE CELLPADDING=4>
  <TR>
    <TD VALIGN=TOP><IMG ALT="Note!" SRC="note.gif"></TD>
    <TD>

<P> There are exceptions when the whole table is not
         scanned. This is when the key part is bound, the key part is
         partially bound in an <CODE>orded_set</CODE> table, or if it is a mnesia
         table and there is a secondary index on the field that is
         selected/matched. Of course if the key is fully bound there will be
         no point in doing a select/match, unless you have a bag table and
         you are only interested in a sub-set of the elements with
         the specific key.
    </TD>
  </TR>
</TABLE>

<P>     When creating a record to be used in a select/match operation you
        want most of the fields to have the value '_'. To avoid having to
        explicitly set all of these fields people have created
        functions that takes advantage of the fact that records are
        implemented as tuples. This is not such a good idea, that is why you
        from R8 can do the following

<PRE>
#person{age = 42, _ = '_'}. 
      
</PRE>

<P>     This will set the age attribute to 42 and all other attributes to '_'.
        This is more efficient then creating a tuple with '_' values, that then is used as
        a record. It is also much better code as it does not
        violate the record abstraction. 
<A NAME="4.1.2"><!-- Empty --></A>
<H4>4.1.2 Deleting an element</H4>

<P>As in the case of lists, the delete operation is considered
        successful if the element was not present in the table. Hence
        all attempts to check that the element is present in the
        Ets/Mnesia table before deletion are unnecessary. Here follows
        an example for Ets tables.
<P><STRONG>DO</STRONG>
<PRE>
...
ets:delete(Tab, Key),
...
      
</PRE>

<P><STRONG>DO NOT</STRONG>
<PRE>
...
case ets:lookup(Tab, Key) of
    [] -&#62;
        ok;
    [_|_] -&#62;
        ets:delete(Tab, Key)
end,
...
      
</PRE>
<A NAME="4.1.3"><!-- Empty --></A>
<H4>4.1.3 Data fetching</H4>

<P>Do not fetch data that you already have! Consider that you
have a module that handles the abstract data type Person. You
export the interface function <CODE>print_person/1</CODE> that uses the internal functions
<CODE>print_name/1</CODE>, <CODE>print_age/1</CODE>, <CODE>print_occupation/1</CODE>.
<P>
<TABLE CELLPADDING=4>
  <TR>
    <TD VALIGN=TOP><IMG ALT="Note!" SRC="note.gif"></TD>
    <TD>

<P>If the functions <CODE>print_name/1</CODE> etc. had been interface
         functions the matter comes in to a whole new light. As you
         do not want the user of the interface to know about the
         internal data representation.     </TD>
  </TR>
</TABLE>

<P><STRONG>DO</STRONG>
<PRE>
%%% Interface function  
print_person(PersonId) -&#62;
    %% Look up the person in the named table person,
    case ets:lookup(person, PersonId) of
        [Person] -&#62;
            print_name(Person),
            print_age(Person),
            print_occupation(Person);
        [] -&#62;
            io:format(&#34;No person with ID = ~p~n&#34;, [PersonID])
    end.

%%% Interanal functions 
print_name(Person) -&#62; 
    io:format(&#34;No person ~p~n&#34;, [Person#person.name]).
                      
print_age(Person) -&#62; 
    io:format(&#34;No person ~p~n&#34;, [Person#person.age]).

print_occupation(Person) -&#62; 
    io:format(&#34;No person ~p~n&#34;, [Person#person.occupation]).

      
</PRE>

<P><STRONG>DO NOT</STRONG>
<PRE>
%%% Interface function          
print_person(PersonId) -&#62;
    %% Look up the person in the named table person,
    case ets:lookup(person, PersonId) of
        [Person] -&#62;
            print_name(PersonID),
            print_age(PersonID),
            print_occupation(PersonID);
        [] -&#62;
            io:format(&#34;No person with ID = ~p~n&#34;, [PersonID])
    end.

%%% Interanal functions         
print_name(PersonID) -&#62; 
    [Person] = ets:lookup(person, PersonId),
    io:format(&#34;No person ~p~n&#34;, [Person#person.name]).
                      
print_age(PersonID) -&#62; 
    [Person] = ets:lookup(person, PersonId),
    io:format(&#34;No person ~p~n&#34;, [Person#person.age]).

print_occupation(PersonID) -&#62; 
    [Person] = ets:lookup(person, PersonId),
    io:format(&#34;No person ~p~n&#34;, [Person#person.occupation]).

      
</PRE>
<A NAME="4.1.4"><!-- Empty --></A>
<H4>4.1.4  Non persistent data storage </H4>

<P>For non persistent database storage, prefer Ets tables before
        Mnesia local_content tables. Even the cheapest Mnesia
        operations, <CODE>dirty_write</CODE> operations, carry a fixed overhead
        compared to Ets writes. Mnesia must check if the table is
        replicated or has indices, this involves at least one Ets
        lookup for each <CODE>dirty_write</CODE>. Thus, Ets writes will always be
        faster than Mnesia writes.<A NAME="4.1.5"><!-- Empty --></A>
<H4>4.1.5 tab2list</H4>

<P>Assume we have an Ets-table, which uses <CODE>idno</CODE> as key,
and contains:
<PRE>
[#person{idno = 1, name = &#34;Adam&#34;,  age = 31, occupation = &#34;mailman&#34;},
 #person{idno = 2, name = &#34;Bryan&#34;, age = 31, occupation = &#34;cashier&#34;},
 #person{idno = 3, name = &#34;Bryan&#34;, age = 35, occupation = &#34;banker&#34;},
 #person{idno = 4, name = &#34;Carl&#34;,  age = 25, occupation = &#34;mailman&#34;}]
      
</PRE>

<P>If we <STRONG>must</STRONG> return all data stored in the Ets-table we
        can use <CODE>ets:tab2list/1</CODE>. However, usually we are only
        interested in a subset of the information in which case
        <CODE>ets:tab2list/1</CODE> is expensive. If we only want to extract
        one field from each record, e.g., the age of every person, we
        should use:
<P><STRONG>DO</STRONG>
<PRE>
...
ets:select(Tab,[{ #person{idno='_', 
                      name='_', 
                      age='$1', 
                      occupation = '_'},
                [],
                ['$1']}]),
...
      
</PRE>

<P><STRONG>DO NOT</STRONG>
<PRE>
...
TabList = ets:tab2list(Tab),
lists:map(fun(X) -&#62; X#person.age end, TabList),
...
      
</PRE>

<P>If we are only interested in the age of all persons named
        Bryan, we should:
<P><STRONG>DO</STRONG>
<PRE>
...
ets:select(Tab,[{ #person{idno='_', 
                      name=&#34;Bryan&#34;, 
                      age='$1', 
                      occupation = '_'},
                [],
                ['$1']}]),
...
      
</PRE>

<P><STRONG>DO NOT</STRONG>
<PRE>
...
TabList = ets:tab2list(Tab),
lists:foldl(fun(X, Acc) -&#62; case X#person.name of
                                &#34;Bryan&#34; -&#62;
                                    [X#person.age|Acc];
                                 _ -&#62;
                                     Acc
                           end
             end, [], TabList),
...
      
</PRE>

<P><STRONG>REALLY DO NOT</STRONG>
<PRE>
...
TabList = ets:tab2list(Tab),
BryanList = lists:filter(fun(X) -&#62; X#person.name == &#34;Bryan&#34; end,
                         TabList),
lists:map(fun(X) -&#62; X#person.age end, BryanList),
...
      
</PRE>

<P>If we need all information stored in the ets table about
        persons named Bryan we should:
<P><STRONG>DO</STRONG>
<PRE>
...
ets:select(Tab, [{#person{idno='_', 
                        name=&#34;Bryan&#34;, 
                        age='_', 
                        occupation = '_'}, [], ['$_']}]),
...
      
</PRE>

<P><STRONG>DO NOT</STRONG>
<PRE>
...
TabList = ets:tab2list(Tab),
lists:filter(fun(X) -&#62; X#person.name == &#34;Bryan&#34; end, TabList),
...
      
</PRE>
<A NAME="4.1.6"><!-- Empty --></A>
<H4>4.1.6 Ordered_set tables</H4>

<P>If the data in the table should be accessed so that the order
        of the keys in the table is significant, the table type
        <CODE>ordered_set</CODE> could be used instead of the more usual
        <CODE>set</CODE> table type. An <CODE>ordered_set</CODE> is always
        traversed in Erlang term order with regards to the key field
        so that return values from functions such as <CODE>select </CODE>,
         <CODE>match_object</CODE> and <CODE>foldl</CODE> are ordered by the key
        values. Traversing an <CODE>ordered_set</CODE> with the first and
        next operations also returns the keys ordered.
<P>
<TABLE CELLPADDING=4>
  <TR>
    <TD VALIGN=TOP><IMG ALT="Note!" SRC="note.gif"></TD>
    <TD>

<P>An <CODE>ordered_set</CODE> only guarantees that
        objects are processed in <STRONG>key</STRONG> order. Results from functions as 
        <CODE>ets:select/2</CODE> appear in the <STRONG>key</STRONG> order even if
        the key is not included in the result.    </TD>
  </TR>
</TABLE>
<A NAME="4.2"><!-- Empty --></A>
<H3>4.2 Ets specific</H3>
<A NAME="4.2.1"><!-- Empty --></A>
<H4>4.2.1 Utilizing the keys of the Ets table</H4>

<P> An Ets table is a singel key table (either a hash table or a
        tree orded by the key) and should be used as one. In other
        words, always use the key to look up things when possible. A
        lookup by a known key in a set Ets table is constant and for a
        orded_set Ets table it is O(logN). A key lookup is always
        preferable to a call where the whole table has to be
        scanned. In the examples above, the field <CODE>idno</CODE> is the
        key of the table and all lookups where only the name is known
        will result in a complete scan of the (possibly large) table
        for a matching result.
<P>A simple solution would be to use the <CODE>name</CODE> field as
        the key instead of the <CODE>idno</CODE> field, but that would cause
        problems if the names were not unique. A more general solution
        would be create a second table with name as key and idno as
        data, i.e. to index (invert) the table with regards to the
        <CODE>name</CODE> field. The second table would of course have to be
        kept consistent with the master table. Mnesia could do this
        for you, but a home brew index table could be very efficient
        compared to the overhead involved in using mnesia.
<P>An index table for the table in the previous examples would
have to be a bag (as keys would appear more than once) and could
have the following contents:
<PRE>
 
[#index_entry{name=&#34;Adam&#34;, idno=1},
 #index_entry{name=&#34;Bryan&#34;, idno=2},
 #index_entry{name=&#34;Bryan&#34;, idno=3},
 #index_entry{name=&#34;Carl&#34;, idno=4}]
      
</PRE>

<P>Given this index table a lookup of the <CODE>age</CODE> fields for
        all persons named &#34;Bryan&#34; could be done like this:
<PRE>
...
MatchingIDs = ets:lookup(IndexTable,&#34;Bryan&#34;),
lists:map(fun(#index_entry{idno = ID}) -&#62;
                 [#person{age = Age}] = ets:lookup(PersonTable, ID),
                 Age
          end,
          MatchingIDs),
...
      
</PRE>

<P>Note that the code above never uses <CODE>ets:match/2</CODE> but
        instead utilizes the <CODE>ets:lookup/2</CODE> call. The
        <CODE>lists:map</CODE> call is only used to traverse the <CODE>idno</CODE>'s
        matching the name &#34;Bryan&#34; in the table, why the number of lookups
        in the master table is minimized.
<P>Keeping an index table of course introduces some overhead when
        inserting records in the table, why the number of operations
        gaining from the table has to be weighted against the number of
        operations inserting objects in the table. However that the gain when
        the key can be used to lookup elements is significant.<A NAME="4.3"><!-- Empty --></A>
<H3>4.3 Mnesia specific</H3>
<A NAME="4.3.1"><!-- Empty --></A>
<H4>4.3.1 Secondary index</H4>

<P>If you frequently do a lookup on a field that is not the
        key of the table, you will lose performance using
        &#34;mnesia:select/match_object&#34; as this function will traverse the
        whole table. You may create a secondary index instead and
        use &#34;mnesia:index_read&#34; to get faster access, however this
        will require more memory. Example:
<PRE>
-record(person, {idno, name, age, occupation}).
        ...
{atomic, ok} = 
mnesia:create_table(person, [{index,[#person.age]},
                              {attributes,
                                    record_info(fields, person)}]),
{atomic, ok} = mnesia:add_table_index(person, age), 
...

PersonsAge42 =
     mnesia:dirty_index_read(person, 42, #person.age),
...
      
</PRE>
<A NAME="4.3.2"><!-- Empty --></A>
<H4>4.3.2  Transactions </H4>

<P>Transactions is a way to guarantee that the distributed
        mnesia database remains consistent, even when many different
        processes updates it in parallel. However if you have
        real time requirements it is recommended to use dirty
        operations instead of transactions. When using the dirty
        operations you lose the consistency guarantee, this is usually
        solved by only letting one process update the table. Other
        processes have to send update requests to that process.
<PRE>
...
% Using transaction
        
Fun = fun() -&#62;
          [mnesia:read({Table, Key}),
           mnesia:read({Table2, Key2})]
          end, 

{atomic, [Result1, Result2]}  = mnesia:transaction(Fun),
...
        
% Same thing using dirty operations
...
        
Result1 = mnesia:dirty_read({Table, Key}),
Result2 = mnesia:dirty_read({Table2, Key2}),
...
      
</PRE>
<A NAME="4.3.3"><!-- Empty --></A>
<H4>4.3.3  Mnemosyne </H4>

<P>Mnesia supports complex queries through the query language
        Mnemosyne. This makes it possible to perform queries of any
        complexity on Mnesia tables. However for simple queries
        Mnemosyne is usually much more expensive than sensible
        handwritten functions doing the same thing.
<P>
<TABLE CELLPADDING=4>
  <TR>
    <TD VALIGN=TOP><IMG ALT="Warning!" SRC="warning.gif"></TD>
    <TD>

<P>The use of
        mnemosyne queries in embedded real time systems is strongly
        discouraged.    </TD>
  </TR>
</TABLE>

<P>Assume we have an mnesia-table, which uses <CODE>idno</CODE> as
key, and contains:
<PRE>
[#person{idno=1, name=&#34;Adam&#34;,  age=31, occupation=&#34;mailman&#34;},
 #person{idno=2, name=&#34;Bryan&#34;, age=31, occupation=&#34;cashier&#34;},
 #person{idno=3, name=&#34;Bryan&#34;, age=35, occupation=&#34;banker&#34;},
 #person{idno=4, name=&#34;Carl&#34;,  age=25, occupation=&#34;mailman&#34;}]
      
</PRE>

<P>If we need to find all persons named Bryan we should:
<P><STRONG>DO </STRONG>
<PRE>
...
Select = fun() -&#62;
                 mnesia:select(person,
                               [{#person{name =&#34;Bryan&#34;, _ = '_'}, [], ['$_']}],
                               read)
         end,
    
{atomic, Result} = mnesia:transaction(Select),
...
      
</PRE>

<P><STRONG>DO NOT</STRONG>
<PRE>
...
Handle = query
         [ Person || Person &#60;- table(person),
                     Person.name = &#34;Bryan&#34;]
        end,
{atomic, Result} = mnesia:transaction(fun() -&#62; mnemosyne:eval(Handle) end),
...
      
</PRE>
<A NAME="4.4"><!-- Empty --></A>
<H3>4.4 Older versions of Erlang/OTP</H3>

<P>If you have a an older version than R8 of Erlang/OTP you would
have to use match and match_object instead of select. The select
call is introduced in R8 and is not present in earlier releases.
Then the code would look as follows.
<P> Selecting the age field: 
<PRE>
...
lists:append(ets:match(Ets, #person{idno='_', 
                                name='_', 
                                age='$1', 
                                occupation = '_'})),
...
    
</PRE>

<P>The <CODE>lists:append/1</CODE> call above transforms the list of
        lists returned by <CODE>ets:match/2</CODE> into a flat list containing
        the values of the field <CODE>age</CODE> in the table.
<P> Selecting people called Bryan: 
<PRE>
...
ets:match_object(Ets, #person{idno='_', 
                          name=&#34;Bryan&#34;, 
                          age='_', 
                          occupation = '_'}),
...
    
</PRE>

<PRE>
...
      
Match = fun() -&#62;
           % Create record instance with '_' as values of the fileds 
           Person = mnesia_table_info(person, wild_pattern),
           mnesia:match_object(person,
                               Person#person{name =&#34;Bryan&#34;},
                               read)
           end,

{atomic, Result} = mnesia:transaction(Match),
...
    
</PRE>
<CENTER>
<HR>
<SMALL>
Copyright &copy; 1991-2006
<A HREF="http://www.erlang.se">Ericsson AB</A><BR>
</SMALL>
</CENTER>
</BODY>
</HTML>