Make GROUP BY work properly for datatypes that only support hashing and not

[PostgreSQL.git] / doc / src / sgml / lobj.sgml

<!-- $PostgreSQL$ -->

 <chapter id="largeObjects">
  <title id="largeObjects-title">Large Objects</title>

  <indexterm zone="largeobjects"><primary>large object</></>
  <indexterm><primary>BLOB</><see>large object</></>

   <para>
    <productname>PostgreSQL</productname> has a <firstterm>large object</>
    facility, which provides stream-style access to user data that is stored
    in a special large-object structure.  Streaming access is useful
    when working with data values that are too large to manipulate
    conveniently as a whole.
   </para>

   <para>
    This chapter describes the implementation and the programming and
    query language interfaces to <productname>PostgreSQL</productname>
    large object data.  We use the <application>libpq</application> C
    library for the examples in this chapter, but most programming
    interfaces native to <productname>PostgreSQL</productname> support
    equivalent functionality.  Other interfaces might use the large
    object interface internally to provide generic support for large
    values.  This is not described here.
   </para>

  <sect1 id="lo-intro">
   <title>Introduction</title>

   <indexterm>
    <primary>TOAST</primary>
    <secondary>versus large objects</secondary>
   </indexterm>

   <para>
    All large objects are placed in a single system table called
    <classname>pg_largeobject</classname>.
    <productname>PostgreSQL</productname> also supports a storage system called
    <quote><acronym>TOAST</acronym></quote> that automatically stores values
    larger than a single database page into a secondary storage area per table.
    This makes the large object facility partially obsolete.  One
    remaining advantage of the large object facility is that it allows values
    up to 2 GB in size, whereas <acronym>TOAST</acronym>ed fields can be at
    most 1 GB.  Also, large objects can be randomly modified using a read/write 
    API that is more efficient than performing such operations using
    <acronym>TOAST</acronym>.
   </para>

  </sect1>

  <sect1 id="lo-implementation">
   <title>Implementation Features</title>

   <para>
    The large object implementation breaks large
    objects up into <quote>chunks</quote> and stores the chunks in
    rows in the database.  A B-tree index guarantees fast
    searches for the correct chunk number when doing random
    access reads and writes.
   </para>
  </sect1>

  <sect1 id="lo-interfaces">
   <title>Client Interfaces</title>

   <para>
    This section describes the facilities that
    <productname>PostgreSQL</productname> client interface libraries
    provide for accessing large objects.  All large object
    manipulation using these functions <emphasis>must</emphasis> take
    place within an SQL transaction block.
    The  <productname>PostgreSQL</productname>  large  object interface is modeled after
    the <acronym>Unix</acronym>  file-system  interface,  with  analogues  of
    <function>open</function>,  <function>read</function>,
    <function>write</function>,
    <function>lseek</function>, etc.
   </para>

   <para>
    Client applications which use the large object interface in
    <application>libpq</application> should include the header file
    <filename>libpq/libpq-fs.h</filename> and link with the
    <application>libpq</application> library.
   </para>

   <sect2>
    <title>Creating a Large Object</title>

    <para>
     The function
<synopsis>
Oid lo_creat(PGconn *conn, int mode);
</synopsis>
     <indexterm><primary>lo_creat</></>
     creates a new large object.  
     The return value is the OID that was assigned to the new large object,
     or <symbol>InvalidOid</symbol> (zero) on failure.

     <replaceable class="parameter">mode</replaceable> is unused and
     ignored as of <productname>PostgreSQL</productname> 8.1; however, for
     backwards compatibility with earlier releases it is best to
     set it to <symbol>INV_READ</symbol>, <symbol>INV_WRITE</symbol>,
     or <symbol>INV_READ</symbol> <literal>|</> <symbol>INV_WRITE</symbol>.
     (These symbolic constants are defined
     in the header file <filename>libpq/libpq-fs.h</filename>.)
    </para>

    <para>
     An example:
<programlisting>
inv_oid = lo_creat(conn, INV_READ|INV_WRITE);
</programlisting>
    </para>

    <para>
     The function
<synopsis>
Oid lo_create(PGconn *conn, Oid lobjId);
</synopsis>
     <indexterm><primary>lo_create</></>
     also creates a new large object.  The OID to be assigned can be
     specified by <replaceable class="parameter">lobjId</replaceable>;
     if so, failure occurs if that OID is already in use for some large
     object.  If <replaceable class="parameter">lobjId</replaceable>
     is <symbol>InvalidOid</symbol> (zero) then <function>lo_create</> assigns an unused
     OID (this is the same behavior as <function>lo_creat</>).
     The return value is the OID that was assigned to the new large object,
     or <symbol>InvalidOid</symbol> (zero) on failure.
    </para>

    <para>
     <function>lo_create</> is new as of <productname>PostgreSQL</productname>
     8.1; if this function is run against an older server version, it will
     fail and return <symbol>InvalidOid</symbol>.
    </para>

    <para>
     An example:
<programlisting>
inv_oid = lo_create(conn, desired_oid);
</programlisting>
    </para>
   </sect2>

   <sect2>
    <title>Importing a Large Object</title>

    <para>
     To import an operating system file as a large object, call
<synopsis>
Oid lo_import(PGconn *conn, const char *filename);
</synopsis>
     <indexterm><primary>lo_import</></>
     <replaceable class="parameter">filename</replaceable> 
     specifies the operating system name of
     the file to be imported as a large object.
     The return value is the OID that was assigned to the new large object,
     or <symbol>InvalidOid</symbol> (zero) on failure.
     Note that the file is read by the client interface library, not by
     the server; so it must exist in the client file system and be readable
     by the client application.
    </para>

    <para>
     The function
<synopsis>
Oid lo_import_with_oid(PGconn *conn, const char *filename, Oid lobjId);
</synopsis>
     <indexterm><primary>lo_import_with_oid</></>
     also imports a new large object.  The OID to be assigned can be
     specified by <replaceable class="parameter">lobjId</replaceable>;
     if so, failure occurs if that OID is already in use for some large
     object.  If <replaceable class="parameter">lobjId</replaceable>
     is <symbol>InvalidOid</symbol> (zero) then <function>lo_import_with_oid</> assigns an unused
     OID (this is the same behavior as <function>lo_import</>).
     The return value is the OID that was assigned to the new large object,
     or <symbol>InvalidOid</symbol> (zero) on failure.
    </para>

    <para>
     <function>lo_import_with_oid</> is new as of <productname>PostgreSQL</productname>
     8.4 and uses <function>lo_create</function> internally which is new in 8.1; if this function is run against 8.0 or before, it will
     fail and return <symbol>InvalidOid</symbol>.
    </para>
   </sect2>

   <sect2>
    <title>Exporting a Large Object</title>

    <para>
     To export a large object
     into an operating system file, call
<synopsis>
int lo_export(PGconn *conn, Oid lobjId, const char *filename);
</synopsis>
     <indexterm><primary>lo_export</></>
     The <parameter>lobjId</parameter> argument specifies the OID of the large
     object to export and the <parameter>filename</parameter> argument
     specifies the operating system name of the file.  Note that the file is
     written by the client interface library, not by the server.  Returns 1
     on success, -1 on failure.
    </para>
   </sect2>

   <sect2>
    <title>Opening an Existing Large Object</title>

    <para>
     To open an existing large object for reading or writing, call
<synopsis>
int lo_open(PGconn *conn, Oid lobjId, int mode);
</synopsis>
     <indexterm><primary>lo_open</></>
     The <parameter>lobjId</parameter> argument specifies the OID of the large
     object to open.   The <parameter>mode</parameter> bits control whether the
     object is opened for reading (<symbol>INV_READ</>), writing
     (<symbol>INV_WRITE</symbol>), or both.
     (These symbolic constants are defined
     in the header file <filename>libpq/libpq-fs.h</filename>.)
     A large object cannot be opened before it is created.
     <function>lo_open</function> returns a (non-negative) large object
     descriptor for later use in <function>lo_read</function>,
     <function>lo_write</function>, <function>lo_lseek</function>,
     <function>lo_tell</function>, and <function>lo_close</function>.
     The descriptor is only valid for 
     the duration of the current transaction.
     On failure, -1 is returned.
    </para>

    <para>
     The server currently does not distinguish between modes
     <symbol>INV_WRITE</symbol> and <symbol>INV_READ</> <literal>|</>
     <symbol>INV_WRITE</symbol>: you are allowed to read from the descriptor
     in either case.  However there is a significant difference between
     these modes and <symbol>INV_READ</> alone: with <symbol>INV_READ</>
     you cannot write on the descriptor, and the data read from it will
     reflect the contents of the large object at the time of the transaction
     snapshot that was active when <function>lo_open</> was executed,
     regardless of later writes by this or other transactions.  Reading
     from a descriptor opened with <symbol>INV_WRITE</symbol> returns
     data that reflects all writes of other committed transactions as well
     as writes of the current transaction.  This is similar to the behavior
     of <literal>SERIALIZABLE</> versus <literal>READ COMMITTED</> transaction
     modes for ordinary SQL <command>SELECT</> commands.
    </para>

    <para>
     An example:
<programlisting>
inv_fd = lo_open(conn, inv_oid, INV_READ|INV_WRITE);
</programlisting>
    </para>
</sect2>

<sect2>
<title>Writing Data to a Large Object</title>

<para>
     The function
<synopsis>
int lo_write(PGconn *conn, int fd, const char *buf, size_t len);
</synopsis>
     <indexterm><primary>lo_write</></> writes
     <parameter>len</parameter> bytes from <parameter>buf</parameter>
     to large object descriptor <parameter>fd</>.  The <parameter>fd</parameter>
     argument must have been returned by a previous
     <function>lo_open</function>.  The number of bytes actually
     written is returned.  In the event of an error, the return value
     is negative.
</para>
</sect2>

<sect2>
<title>Reading Data from a Large Object</title>

<para>
     The function
<synopsis>
int lo_read(PGconn *conn, int fd, char *buf, size_t len);
</synopsis>
     <indexterm><primary>lo_read</></> reads
     <parameter>len</parameter> bytes from large object descriptor
     <parameter>fd</parameter> into <parameter>buf</parameter>. The
     <parameter>fd</parameter> argument must have been returned by a
     previous <function>lo_open</function>.  The number of bytes
     actually read is returned. In the event of an error, the return
     value is negative.
</para>
</sect2>

<sect2>
<title>Seeking in a Large Object</title>

<para>
     To change the current read or write location associated with a
     large object descriptor, call
<synopsis>
int lo_lseek(PGconn *conn, int fd, int offset, int whence);
</synopsis>
     <indexterm><primary>lo_lseek</></> This function moves the
     current location pointer for the large object descriptor identified by
     <parameter>fd</> to the new location specified by
     <parameter>offset</>.  The valid values for <parameter>whence</>
     are <symbol>SEEK_SET</> (seek from object start),
     <symbol>SEEK_CUR</> (seek from current position), and
     <symbol>SEEK_END</> (seek from object end).  The return value is
     the new location pointer, or -1 on error.
</para>
</sect2>

<sect2>
<title>Obtaining the Seek Position of a Large Object</title>

<para>
     To obtain the current read or write location of a large object descriptor,
     call
<synopsis>
int lo_tell(PGconn *conn, int fd);
</synopsis>
     <indexterm><primary>lo_tell</></> If there is an error, the
     return value is negative.
</para>
</sect2>

<sect2>
<title>Truncating a Large Object</title>

<para>
     To truncate a large object to a given length, call
<synopsis>
int lo_truncate(PGcon *conn, int fd, size_t len);
</synopsis>
     <indexterm><primary>lo_truncate</></> truncates the large object
     descriptor <parameter>fd</> to length <parameter>len</>.  The
     <parameter>fd</parameter> argument must have been returned by a
     previous <function>lo_open</function>.  If <parameter>len</> is
     greater than the current large object length, the large object
     is extended with null bytes ('\0').
</para>

<para>
     The file offset is not changed.
</para>

<para>
     On success <function>lo_truncate</function> returns
     zero.  On error, the return value is negative.
</para>

<para>
     <function>lo_truncate</> is new as of <productname>PostgreSQL</productname>
     8.3; if this function is run against an older server version, it will
     fail and return a negative value.
</para>
</sect2>

<sect2>
<title>Closing a Large Object Descriptor</title>

<para>
     A large object descriptor can be closed by calling
<synopsis>
int lo_close(PGconn *conn, int fd);
</synopsis>
     <indexterm><primary>lo_close</></> where <parameter>fd</> is a
     large object descriptor returned by <function>lo_open</function>.
     On success, <function>lo_close</function> returns zero.  On
     error, the return value is negative.
</para>

<para>
     Any large  object  descriptors that remain open at the end of a
     transaction will be closed automatically.
</para>
</sect2>

   <sect2>
    <title>Removing a Large Object</title>

    <para>
     To remove a large object from the database, call
<synopsis>
int lo_unlink(PGconn *conn, Oid lobjId);
</synopsis>
     <indexterm><primary>lo_unlink</></> The
     <parameter>lobjId</parameter> argument specifies the OID of the
     large object to remove.  Returns 1 if successful, -1 on failure.
    </para>
   </sect2>

</sect1>

<sect1 id="lo-funcs">
<title>Server-Side Functions</title>

  <para>
   There are server-side functions callable from SQL that correspond to
   each of the client-side functions described above; indeed, for the
   most part the client-side functions are simply interfaces to the
   equivalent server-side functions.  The ones that are actually useful
   to call via SQL commands are
   <function>lo_creat</function><indexterm><primary>lo_creat</></>,
   <function>lo_create</function><indexterm><primary>lo_create</></>,
   <function>lo_unlink</function><indexterm><primary>lo_unlink</></>,
   <function>lo_import</function><indexterm><primary>lo_import</></>, and
   <function>lo_export</function><indexterm><primary>lo_export</></>.
   Here are examples of their use:

<programlisting>
CREATE TABLE image (
    name            text,
    raster          oid
);

SELECT lo_creat(-1);       -- returns OID of new, empty large object

SELECT lo_create(43213);   -- attempts to create large object with OID 43213

SELECT lo_unlink(173454);  -- deletes large object with OID 173454

INSERT INTO image (name, raster)
    VALUES ('beautiful image', lo_import('/etc/motd'));

INSERT INTO image (name, raster)  -- same as above, but specify OID to use
    VALUES ('beautiful image', lo_import('/etc/motd', 68583));

SELECT lo_export(image.raster, '/tmp/motd') FROM image
    WHERE name = 'beautiful image';
</programlisting>
  </para>

  <para>
    The server-side <function>lo_import</function> and
    <function>lo_export</function> functions behave considerably differently
    from their client-side analogs.  These two functions read and write files
    in the server's file system, using the permissions of the database's
    owning user.  Therefore, their use is restricted to superusers.  In
    contrast, the client-side import and export functions read and write files
    in the client's file system, using the permissions of the client program.
    The client-side functions can be used by any
    <productname>PostgreSQL</productname> user.
  </para>
</sect1>

<sect1 id="lo-examplesect">
<title>Example Program</title>

<para>
     <xref linkend="lo-example"> is a sample program which shows how the large object  
     interface
     in <application>libpq</> can be used.  Parts of the program are 
     commented out but are left in the source for  the  reader's
     benefit.  This program can also be found in
     <filename>src/test/examples/testlo.c</filename> in the source distribution.
</para>

  <example id="lo-example">
   <title>Large Objects with <application>libpq</application> Example Program</title>
<programlisting>
/*--------------------------------------------------------------
 *
 * testlo.c--
 *    test using large objects with libpq
 *
 * Copyright (c) 1994, Regents of the University of California
 *
 *--------------------------------------------------------------
 */
#include &lt;stdio.h&gt;
#include &quot;libpq-fe.h&quot;
#include &quot;libpq/libpq-fs.h&quot;

#define BUFSIZE          1024

/*
 * importFile
 *    import file &quot;in_filename&quot; into database as large object &quot;lobjOid&quot;
 *
 */
Oid
importFile(PGconn *conn, char *filename)
{
    Oid         lobjId;
    int         lobj_fd;
    char        buf[BUFSIZE];
    int         nbytes,
                tmp;
    int         fd;

    /*
     * open the file to be read in
     */
    fd = open(filename, O_RDONLY, 0666);
    if (fd &lt; 0)
    {                           /* error */
        fprintf(stderr, &quot;cannot open unix file %s\n&quot;, filename);
    }

    /*
     * create the large object
     */
    lobjId = lo_creat(conn, INV_READ | INV_WRITE);
    if (lobjId == 0)
        fprintf(stderr, &quot;cannot create large object\n&quot;);

    lobj_fd = lo_open(conn, lobjId, INV_WRITE);

    /*
     * read in from the Unix file and write to the inversion file
     */
    while ((nbytes = read(fd, buf, BUFSIZE)) &gt; 0)
    {
        tmp = lo_write(conn, lobj_fd, buf, nbytes);
        if (tmp &lt; nbytes)
            fprintf(stderr, &quot;error while reading large object\n&quot;);
    }

    (void) close(fd);
    (void) lo_close(conn, lobj_fd);

    return lobjId;
}

void
pickout(PGconn *conn, Oid lobjId, int start, int len)
{
    int         lobj_fd;
    char       *buf;
    int         nbytes;
    int         nread;

    lobj_fd = lo_open(conn, lobjId, INV_READ);
    if (lobj_fd &lt; 0)
    {
        fprintf(stderr, &quot;cannot open large object %d\n&quot;,
                lobjId);
    }

    lo_lseek(conn, lobj_fd, start, SEEK_SET);
    buf = malloc(len + 1);

    nread = 0;
    while (len - nread &gt; 0)
    {
        nbytes = lo_read(conn, lobj_fd, buf, len - nread);
        buf[nbytes] = ' ';
        fprintf(stderr, &quot;&gt;&gt;&gt; %s&quot;, buf);
        nread += nbytes;
    }
    free(buf);
    fprintf(stderr, &quot;\n&quot;);
    lo_close(conn, lobj_fd);
}

void
overwrite(PGconn *conn, Oid lobjId, int start, int len)
{
    int         lobj_fd;
    char       *buf;
    int         nbytes;
    int         nwritten;
    int         i;

    lobj_fd = lo_open(conn, lobjId, INV_WRITE);
    if (lobj_fd &lt; 0)
    {
        fprintf(stderr, &quot;cannot open large object %d\n&quot;,
                lobjId);
    }

    lo_lseek(conn, lobj_fd, start, SEEK_SET);
    buf = malloc(len + 1);

    for (i = 0; i &lt; len; i++)
        buf[i] = 'X';
    buf[i] = ' ';

    nwritten = 0;
    while (len - nwritten &gt; 0)
    {
        nbytes = lo_write(conn, lobj_fd, buf + nwritten, len - nwritten);
        nwritten += nbytes;
    }
    free(buf);
    fprintf(stderr, &quot;\n&quot;);
    lo_close(conn, lobj_fd);
}

/*
 * exportFile
 *    export large object &quot;lobjOid&quot; to file &quot;out_filename&quot;
 *
 */
void
exportFile(PGconn *conn, Oid lobjId, char *filename)
{
    int         lobj_fd;
    char        buf[BUFSIZE];
    int         nbytes,
                tmp;
    int         fd;

    /*
     * open the large object
     */
    lobj_fd = lo_open(conn, lobjId, INV_READ);
    if (lobj_fd &lt; 0)
    {
        fprintf(stderr, &quot;cannot open large object %d\n&quot;,
                lobjId);
    }

    /*
     * open the file to be written to
     */
    fd = open(filename, O_CREAT | O_WRONLY, 0666);
    if (fd &lt; 0)
    {                           /* error */
        fprintf(stderr, &quot;cannot open unix file %s\n&quot;,
                filename);
    }

    /*
     * read in from the inversion file and write to the Unix file
     */
    while ((nbytes = lo_read(conn, lobj_fd, buf, BUFSIZE)) &gt; 0)
    {
        tmp = write(fd, buf, nbytes);
        if (tmp &lt; nbytes)
        {
            fprintf(stderr, &quot;error while writing %s\n&quot;,
                    filename);
        }
    }

    (void) lo_close(conn, lobj_fd);
    (void) close(fd);

    return;
}

void
exit_nicely(PGconn *conn)
{
    PQfinish(conn);
    exit(1);
}

int
main(int argc, char **argv)
{
    char       *in_filename,
               *out_filename;
    char       *database;
    Oid         lobjOid;
    PGconn     *conn;
    PGresult   *res;

    if (argc != 4)
    {
        fprintf(stderr, &quot;Usage: %s database_name in_filename out_filename\n&quot;,
                argv[0]);
        exit(1);
    }

    database = argv[1];
    in_filename = argv[2];
    out_filename = argv[3];

    /*
     * set up the connection
     */
    conn = PQsetdb(NULL, NULL, NULL, NULL, database);

    /* check to see that the backend connection was successfully made */
    if (PQstatus(conn) == CONNECTION_BAD)
    {
        fprintf(stderr, &quot;Connection to database '%s' failed.\n&quot;, database);
        fprintf(stderr, &quot;%s&quot;, PQerrorMessage(conn));
        exit_nicely(conn);
    }

    res = PQexec(conn, &quot;begin&quot;);
    PQclear(res);

    printf(&quot;importing file %s\n&quot;, in_filename);
/*  lobjOid = importFile(conn, in_filename); */
    lobjOid = lo_import(conn, in_filename);
/*
    printf(&quot;as large object %d.\n&quot;, lobjOid);

    printf(&quot;picking out bytes 1000-2000 of the large object\n&quot;);
    pickout(conn, lobjOid, 1000, 1000);

    printf(&quot;overwriting bytes 1000-2000 of the large object with X's\n&quot;);
    overwrite(conn, lobjOid, 1000, 1000);
*/

    printf(&quot;exporting large object to file %s\n&quot;, out_filename);
/*    exportFile(conn, lobjOid, out_filename); */
    lo_export(conn, lobjOid, out_filename);

    res = PQexec(conn, &quot;end&quot;);
    PQclear(res);
    PQfinish(conn);
    exit(0);
}
</programlisting>
</example>

</sect1>
</chapter>