libpqtypes home page

pqt-specs(3)                   libpqtypes Manual                  pqt-specs(3)



NAME
       pqt-specs - A manual for libpqtypes data type specifier strings.

DESCRIPTION
       The  type  system  allows  addressing backend data types by their fully
       qualified schema name.  Similar to the backend type system,  libpqtypes
       has  an  input  and  output function for each type: put and get respec-
       tively.  All builtin types are  supported  by  libpqtypes.   Additional
       types,  such as composites and user-defined types, can be registered by
       an API-user on a per connection basis.

       Putting and getting types are  addressed  by  their  backend  names  as
       printf-style  format  specifiers.  For instance: %int4, %timestamptz or
       %text.  They can also be addressed using their fully  qualified  schema
       names:  %pg_catalog.int4,  %pg_catalog.timestamptz, %pg_catalog.text or
       %myschema.mytype to avoid ambiguity.


   Specifier Strings
       Specifier Mark
       A specifier mark indicates where a specifier begins. A mark is not part
       of the type name.

          %  Marks the beginning of a type name.  When getting results, this
             also indicates that a field will be referenced by number.

          #  Marks the beginning of a type name.  When getting results, this
             also indicates that a field will be referenced by name.

          @  Marks the beginning of a prepared specifier name, see
             PQspecPrepare().  When used, it must be the first and only
             specifier in the format string: "@prepared_spec", "@myfunc",
             etc...  NOTE: the '@' must be the first character, no spaces.

       Type Specifier
       Type specifiers are comprised of an optional schema name and type name.
       Type specifiers have a set of rules:

          -) Format: [schema].type - optional schema name, a "." separator
             between schema and type and the type name.

          -) First character must be a-z or an underscore

          -) Double quotes are required for characters not in [a-zA-Z0-9_]
             NOTE: In libpqtypes, this includes "double precision"

          -) Schema "." separator, specifier marks or flags are not included
             in double quotes

          -) Non-quoted type names are casefolded, quoted names are not.

          -) Examples:
             "%int4"
             "%pg_catalog.int4"
             "%\"my oDd~ !tYpe naMe#\""
             "%myschema.\"my oDd~ !tYpe naMe#\""
             "%abc.int4 %pg_catalog.int4" <= fully qualified type names

          -) Last in First out: To find a type referenced in a specifier
             string, the search begins with the last type registered.
             User registered type handlers are searched first, followed
             by builtin types.

          -) pqt schema: There is a builtin schema named pqt.  By default,
             it contains two types: 'str' and 'null'.  Anything can be
             put into this schema, which has nothing to do with the server.
             This is good for aliases or type sub-classes that are
             client-specific.

       Specifier Flag
       Flags are used to alter the behavior of a  type  specifier.   They  are
       always  placed at the end of the specifier name.  If the name is double
       quoted, the flag is just after the closing quote.

          *  This is called the pointer flag.  It is only supported on a
             handful of builtin types during a PQputf(3), but user registered
             types can provide support for them.  Supported types are:
             VARCHAR, BPCHAR, TEXT, BYTEA, NAME and the pqt.str.

             Putting data: this flag tells libpqtypes to store a direct
             pointer to the data being put, rather than making a copy of
             it.

             Getting data: no built-in types make use of the pointer flag.
             User-defined type handlers can make the pointer flag behave
             anyway they see fit.  The 'get' type handler is supplied a
             PGtypeArgs which contains an 'is_ptr' member.

          [] This is called the array flag.  It indicates that an array is
             being referenced rather than a simple type.  This flag is always
             used with a PGarray.

NUMERICS
       With the exception of the "numeric"  type,  all  numeric  types  behave
       identically: int2, int4, int8, float4 and float8.

              PG type    C type
              PGint2     short
              PGint4     int
              PGint8     long long (platform dependent)
              PGfloat4   float
              PGfloat8   double

       Putting  numeric  values: If the value supplied is too large for the PG
       type, it will be silently truncated.

              PQputf(param, "%int2 %int4 %int8 %float4 %float8",
                   SHRT_MAX, INT_MAX, LLONG_MAX, 1234.56, 123456.789);

       Getting numeric values: Like scanf, the correctly sized data type  must
       be  used.   For  instance:  you cannot use a 4-byte int for %int2 - you
       must use a short.

              // Read an int2 from field 0, int4 from field 1, int8 from
              // field 2, float4 from field 3 and a float8 from field 4
              PGint2 i2;
              PGint4 i4;
              PGint8 i8;
              PGfloat4 f4;
              PGfloat8 f8;
              PQgetf(result, tup_num, "%int2 %int4 %int8 %float4 %float8",
                   0, &i2, 1, &i4, 2, &i8, 3, &f4, 4, &f8);

       The numeric data type is always exposed in text format.  There is no  C
       data structure. It is always sent in binary format.

              PQputf(param, "%numeric", "1728718718271827121233.121212121212");

       Even  if  binary  results are used when getting a numeric value, libpq-
       types will internally convert the numeric to text. This has the  advan-
       tage  of  allowing  you  to use binary results and still have access to
       numeric fields.  If you want to work with a numeric in binary form, use
       PQgetvalue() on a binary result set.

              PGnumeric numstr;
              PQgetf(result, tup_num, "%numeric", field_num, &numstr);

       The first argument is the field number of the numeric. The second argu-
       ment is a pointer to a PGnumeric to receive the  numeric  string  value
       (which will always be NUL terminated).

ARRAY
       Arrays  are  put  using  the PGarray structure.  Elements are put using
       PQputf(3) into a PGparam structure  contained  withn  a  PGarray.   The
       PGarray  contains  array  dimension members for specifing the number of
       dimension, the dimensions of each dimension and  the  lower  bounds  of
       each dimension.

       Arrays  are only handled using binary format.  This means that any type
       used as an array element must be put and gotten in binary format.  If a
       user-defined  type  does  not implement a send and recv function in the
       backend, it can not be used as an array element.

       For a discussion of composite arrays, `man pqt-composites(3)'.

              typedef struct
              {
                   /* The number of array dimensions.  Specifing zero for this
                    * value on puts has special meaning.  When zero, this value
                    * is set to one, dims[0] is set to the number of items in
                    * the 'param' member and lbound[0] is set to one.
                    */
                   int ndims;

                   /* An array of lower bounds for each dimension. */
                   int lbound[MAXDIM];

                   /* An array of dimensions for each dimension. */
                   int dims[MAXDIM];

                   /* When putting array elements, this PGparam is used.  Each
                    * item put is one array element.  Because the PQputf(3)
                    * interface allows putting more than one item at a time, you
                    * can put multiple array elements.
                    */
                   PGparam *param;

                   /* When getting an array, this is the PGresult object that
                    * contains the array elements.  Each element is one tuple,
                    * regardless of the array dimensions.  If the array has 100
                    * elements across 3 dimensions, PQntuples(arr.res) will return
                    * 100.  The only valid field, for non-composite arrays, is
                    * field zero.
                    */
                   PGresult *res;
              } PGarray;

       When all elements have been put, the  PGarray  structure  must  be  put
       using the "[]" array specifer flag into a PGparam structure.  PQputf(3)
       is used to build the array elements and to put the resulting PGarray.

       Putting an array value:

              PGint4 i;
              PGarray arr;
              PGparam *param;

              /* One dimensional arrays do not require setting dimension info. For
               * convience, you can zero the structure or set ndims to zero.
               */
              arr.ndims = 0;

              /* create the param object that will contain the elements */
              arr.param = PQparamCreate(conn);

              /* Use PQputf(3) to put the array elements */
              for(i=0; i < 1000; i++)
                PQputf(arr.param, "%int4", i);

              /* The PGarray must be put into a PGparam struture.  So far, only
               * the array elements have been put.  'param' can continue to be
               * used to pack more parameters.  The array is now a single parameter
               * within 'param'.
               */
              param = PQparamCreate(conn);
              PQputf(param, "%int[]", &arr);

              /* no longer needed */
              PQparamClear(arr.param);

       To get an array, PQgetf(3) is used  in  conjunction  with  the  PGarray
       structure.   The  result object contained with the PGarray is populated
       with the array elements.  The dimension info is assigned as well.  Each
       array  element  is  its own tuple that only contains a single field for
       non composite arrays.

       Getting an array value:

              int i;
              PGint4 val;
              int ntups;
              PGarray arr;
              PGresult *result = ...;

              PQgetf(result, 0, "%int[]", 0, &arr);

              /* not needed anymore*/
              PQclear(result);

              /* each tuple is an array element */
              ntups = PQntuples(arr.res);
              for(i=0; i < ntups; i++)
              {
                /* Always field 0 */
                PQgetf(arr.res, i, "%int4", 0, &val);
                printf("[%03d] %d\n", i, val);
              }

              PQclear(arr.res);

       The result object  is  not  organized  based  on  the  dimension  info.
       Indexes  are always zero-based.  If the dimension info is meaningful to
       your application, index translation must be done using the ndims,  dims
       and lbound members of the PGarray structure.

       You  cannot reference an array element by field name.  The only logical
       name for an array element would be the string version of its index  ...
       "0",  "1", etc..  The index value of a non-composite array is its tuple
       number, the field number is always zero.   This  means  "#int"  is  not
       legal  for  non-composite  arrays.   You  must use "%int" and reference
       field 0.

CHAR
       The "char" data type uses the PGchar.  The value is limited to  8-bits.

       Putting a "char" value:

              PGchar c = 'a';
              PQputf(param, "%char %char", 213, c);

       Getting a "char" value:

              PGchar c;
              PQgetf(result, tup_num, "%char", field_num, &c);

VARIABLE-LENGTH CHARACTER TYPES
       BPCHAR,  VARCHAR,  NAME  and  TEXT are handled identically.  libpqtypes
       does no range checking on NAME, BPCHAR or VARCHAR, it lets  the  server
       perform  these checks.  There are two ways to put strings: allow libpq-
       types to make an internal copy of the string (default behavior) or as a
       direct pointer: (both require that the C string is NUL-terminated)

              /* Put a string so libpqtypes makes a copy. In this case,
               * 4 copies would be made of the same string.
               */
              PGtext str = "foobar";
              PQputf(param, "%bpchar %varchar %name %text", str, str, str, str);

              /* Put a string so libpqtypes doesn't make a copy,
               * keeps a direct pointer.  More efficient than above,
               * especially if these are large strings.
               */
              PQputf(param, "%bpchar* %varchar* %name* %text*", str, str, str, str);

       WARNING:  Be  careful about variable scope when using the "*" specifier
       flag:

              /* when 'func' returns, the str pointer becomes invalid!
               * The below should be using "%text" ... w/o the * flag.
               */
              int func(PGparam *param)
              {
                   PGchar str[16];
                   strcpy(str, "foobar");
                   return PQputf(param, "%text*", str); // BAD IDEA!
              }

       To PQgetf(3) a string, you  supply  a  pointer  to  a  PGtext.   Unlike
       putting  string values, getting them doesn't make use of the "*" speci-
       fier flag (silently ignored).

              /* Get a string value */
              PGvarchar str;
              PQgetf(result, tup_num, "%varchar", field_num, &str);

              /* identical to */
              str = PQgetvalue(result, tup_num, field_num);

       The reason the '*' specifier flag  is  silently  ignored,  rather  than
       raising  a  syntax error, is it is common to define macros for specifer
       strings; that can be used for puts and gets:

              /* user_id, username, password */
              #define TYPESPEC_USERINFO "%int4 %text* %text*"

              PGint4 uid = 0;
              PGtext user = "foo", pass = "bar";
              PQputf(param, TYPESPEC_USERINFO, uid, user, pass);
              PQgetf(param, tup_num, TYPESPEC_USERINFO, 0, &uid, 1, &user, 2, &pass);

       The above allowance is more useful than a syntax error.

BYTEA
       There are two ways to put a bytea: copy or direct  pointer  (just  like
       variable-length character types).  In either case, you supply a pointer
       to a PGbytea.

              typedef struct
              {
                int len;    /* number of bytes */
                char *data; /* pointer to the bytea data */
              } PGbytea;

              /* Put a bytea letting libpqtypes make a copy; */
              PGbytea bytea = {4, {0, 1, 2, 3}};
              PQputf(param, "%bytea", &bytea);

              /* Put a bytea not letting libpqtypes make a copy, stores a
               * direct pointer to PGbytea.data.
               */
              PQputf(param, "%bytea*", &bytea);

       To get a bytea, you provide a pointer to  a  PGbytea.   Unlike  putting
       bytea values, there is only one way to get them.

              /* Get a bytea value (exposed as binary, no
               * escaping/unescaping needed)
               */
              PGbytea bytea;
              PQgetf(result, tup_num, "%bytea", field_num, &bytea);

       NOTE:  For text results, PQgetlength will not match the length returned
       by PQgetf(3).  The  values  PQgetf(3)  assigns  to  the  user  provided
       PGbytea* represent the unescaped bytea value.

DATE
       PGdate  is used by DATE, TIMESTAMP and TIMESTAMPTZ data types. To put a
       date, you must set the isbc, year, mon and  mday  members.   All  other
       members are ignored.

              typedef struct
              {
                   /* When non-zero, the date is in the BC ERA. */
                   int isbc;

                   /*
                    * The BC or AD year, which is NOT adjusted by 1900 like
                    * the POSIX struct tm.  Years are always positive values,
                    * even BC years.  To distinguish between BC and AD years,
                    * use the isbc flag: (year 0 not used)
                    *   Ex. -1210 is represented as: isbc=1, year=1209
                    */
                   int year;

                   /* The number of months since January, in the range 0 to 11. */
                   int mon;

                   /* The day of the month, in the range 1 to 31. */
                   int mday;

                   /* The Julian day in the Gregorian calendar. */
                   int jday;

                   /* The number of days since January 1, in the range 0 to 365. */
                   int yday;

                   /* The number of days since Sunday, in the range 0 to 6. */
                   int wday;
              } PGdate;

       Putting a date value:

              // '1401-01-19 BC'
              PGdate date;
              date.isbc = 1;
              date.year = 1401;
              date.mon  = 0;
              date.mday = 19;
              PQputf(param, "%date", &date);

       Getting a date value:

              PQgetf(result, tup_num, "%date", field_num, &date);

TIME
       PGtime  is  used by TIME, TIMETZ, TIMESTAMP and TIMESTAMPTZ data types.
       To put a time, you must set the hour, min, sec and usec  members.   All
       other members are ignored.

              typedef struct
              {
                   /* The number of hours past midnight, in the range 0 to 23. */
                   int hour;

                   /* The number of minutes after the hour, in the
                    * range 0 to 59.
                    */
                   int min;

                   /* The number of seconds after the minute, in the
                    * range 0 to 59.
                    */
                   int sec;

                   /* The number of microseconds after the second, in the
                    * range of 0 to 999999.
                    */
                   int usec;

                   /*
                    * When non-zero, this is a TIME WITH TIME ZONE.  Otherwise,
                    * it is a TIME WITHOUT TIME ZONE.
                    */
                   int withtz;

                   /* A value of 1 indicates daylight savings time.  A value of 0
                    * indicates standard time.  A value of -1 means unknown or
                    * could not determine.
                    */
                   int isdst;

                   /* Offset from UTC in seconds. This value is not always
                    * available. It is set to 0 if it cannot be determined.
                    */
                   int gmtoff;

                   /* Timezone abbreviation: such as EST, GMT, PDT, etc.
                    * This value is not always available.  It is set to an empty
                    * string if it cannot be determined.
                    */
                   char tzabbr[16];
              } PGtime;

       Putting a time value:

              // '10:41:06.002897'
              PGdate time;
              time.hour   = 10;
              time.min    = 41;
              time.sec    = 6;
              time.usec   = 2897;
              PQputf(param, "%time", &time);

       Getting a time value:

              PQgetf(result, tup_num, "%time", field_num, &time);

TIMETZ
       The  TIMETZ  data  type uses the PGtime structure, for a description of
       this structure see the TIME section.  To put a timetz, you must set the
       hour,  min,  sec,  usec  and  gmtoff  members.   All  other members are
       ignored.

       Putting a timetz value:

              // '10:41:06.002897-05'
              PGdate timetz;
              timetz.hour   = 10;
              timetz.min    = 41;
              timetz.sec    = 6;
              timetz.usec   = 2897;
              timetz.gmtoff = -18000;
              PQputf(param, "%timetz", &timetz);

       Getting a timetz value:

              PQgetf(result, tup_num, "%timetz", field_num, &timetz);

TIMESTAMP
       To put a timestamp, the isbc, year, mon, mday, hour, min, sec and  usec
       members must be set.  No other members are used.

              typedef struct
              {
                   /* The number seconds before or after midnight UTC of
                    * January 1, 1970, not counting leap seconds.
                    */
                   PGint8 epoch;

                   /* The date part of the timestamp. */
                   PGdate date;

                   /* The time part of the timestamp. */
                   PGtime time;
              } PGtimestamp;

       Putting a timestamp value:

              // '2000-01-19 10:41:06'
              PGtimestamp ts;
              ts.date.isbc   = 0;
              ts.date.year   = 2000;
              ts.date.mon    = 0;
              ts.date.mday   = 19;
              ts.time.hour   = 10;
              ts.time.min    = 41;
              ts.time.sec    = 6;
              ts.time.usec   = 0;
              PQputf(param, "%timestamp", &ts);

       Getting a timestamp value:

              PQgetf(result, tup_num, "%timestamp", field_num, &ts);

       The timestamp type has no concept of timezone, so the value returned by
       PQgetf(3) is exactly what the server sent; no timezone adjustments  are
       attempted.   The  gmtoff is always set to zero, tzabbr will be an empty
       string and withtz will be zero.

TIMESTAMPTZ
       To put a timestamptz, the isbc, year, mon, mday, hour, min,  sec,  usec
       and gmtoff members must be set.  No other members are used.

       Putting a timestamptz value:

              // '2000-01-19 10:41:06-05'
              PGtimestamp ts;
              ts.date.isbc   = 0;
              ts.date.year   = 2000;
              ts.date.mon    = 0;
              ts.date.mday   = 19;
              ts.time.hour   = 10;
              ts.time.min    = 41;
              ts.time.sec    = 6;
              ts.time.usec   = 0;
              ts.time.gmtoff = -18000;
              PQputf(param, "%timestamptz", &ts);

       Getting a timestamptz value:

              PQgetf(result, tup_num, "%timestamptz", field_num, &ts);

       When using PQgetf(3) in binary mode, the timestamptz value is converted
       into the local machine's timezone.  If the local machine's timezone can
       not  be determined, the value will be in GMT (gmtoff is set to zero and
       tzabbr is set to GMT).

       When using PQgetf(3) in text mode, the timestamptz value is returned as
       a datetime string in the server's timezone.  No adjustments are made to
       this value.  If the server is using a DateStyle that encodes the gmtoff
       "00:00:00-05", then gmtoff will be set to this value and tzabbr will be
       "GMT+/-hhmmss" (00:00:00-05 => GMT-0500).  In this case, isdst  is  set
       to  -1  ... meaning unknown.  If the server's DateStyle encodes a time-
       zone abbreviation, like PST, then tzabbr is set  to  this  value.   The
       gmtoff and isdst members are properly set:

         DateStyle includes a timezone abbrev - "SQL, MDY"
         01/25/2007 00:00:00 EST => tzabbr=EST, gmtoff=-18000, isdst=0
         01/25/2007 01:00:00 EDT => tzabbr=EDT, gmtoff=-14400, isdst=1

INTERVAL
       To  put  an  interval,  all  relevant members of a PGinterval should be
       assigned and those not used should be set to zero.

              typedef struct
              {
                   /* the number of years */
                   int years;

                   /* the number of months */
                   int mons;

                   /* the number of days */
                   int days;

                   /* the number of hours */
                   int hours;

                   /* the number of mins */
                   int mins;

                   /* the number of seconds */
                   int secs;

                   /* the number of microseconds */
                   int usecs;
              } PGinterval;

       Putting an interval value:

              // "20 years 8 months 9 hours 10 mins 15 secs 123456 usecs"
              PGinterval interval;
              interval.years = 20;
              interval.mons  = 8;
              interval.days  = 0; // not used, set to 0
              interval.hours = 9;
              interval.mins  = 10;
              interval.secs  = 15;
              interval.usecs = 123456;
              PQputf(param, "%interval", &interval);

       Getting an interval value:

              PQgetf(result, tup_num, "%interval", field_num, &interval);

       NOTE: When using text results with a  non-ISO  DateStyle,  microseconds
       are  truncated to a 2 digit value.  For example: "4 mins 2.11 secs" but
       microseconds is really 111456.

POINT
       The PGpoint structure is used to put and get a point.

              typedef struct
              {
                   double x; // point x value
                   double y; // point y value
              } PGpoint;

       Putting a point value:

              PGpoint pt = {12.345, 6.789};
              PQputf(param, "%point", &pt);

       Getting a point value:

              PGpoint pt;
              PQgetf(result, tup_num, "%point", field_num, &pt);

LSEG
       The PGlseg structure is used to put and get a line segnment.

              typedef struct
              {
                PGpoint pts[2];
              } PGlseg;

       Putting a lseg value:

              PGlseg lseg = {{{12.345, 6.789}, {99.8, 88.9}}};
              PQputf(param, "%lseg", &lseg);

       Getting a lseg value:

              PGlseg lseg;
              PQgetf(result, tup_num, "%lseg", field_num, &lseg);

BOX
       The PGbox structure is used to put and get a box.

              typedef struct
              {
                PGpoint high;
                PGpoint low;
              } PGbox;

       Putting a box value:

              PGbox box = {{12.345, 6.789}, {22.234, 1.9998}};
              PQputf(param, "%box", &box);

       Getting a box value:

              PGbox box;
              PQgetf(result, tup_num, "%box", field_num, &box);

CIRCLE
       The PGcircle structure is used to put and get a circle.

              typedef struct
              {
                PGpoint center;
                double radius;
              } PGcircle;

       Putting a circle value:

              PGcircle circle = {{12.345, 6.789}, 2.34567};
              PQputf(param, "%circle", &circle);

       Getting a circle value:

              PGcircle circle;
              PQgetf(result, tup_num, "%circle", field_num, &circle);

PATH
       The PGpath structure is used to put and get a path.  If the closed mem-
       ber is non-zero, the path is closed, otherwise it is open.

              typedef struct
              {
                int npts;
                int closed;
                PGpoint *pts;
              } PGpath;

       Putting a path value:

              // Put a closed path that contains 2 points
              PGpoint pts[] = {{12.345, 6.789}, {19.773, 7.882}};
              PGpath path = {2, 1, pts};
              PQputf(param, "%path", &path);

       Getting a path value:

              PGpath path;
              if(PQgetf(result, tup_num, "%path", field_num, &path))
              {
                   // path.pts must be copied out if needed after clearing results
                   copy_points(path.npts, path.pts, ...);

                   PQclear(result);
                   // path.pts is now invalid!
              }

POLYGON
       The PGpolygon structure is used to put and get a polygon.

              typedef struct
              {
                int npts;
                PGpoint *pts;
              } PGpolygon;

       Putting a polygon value:

              // Put a polygon that contains 2 points
              PGpoint pts[] = {{12.345, 6.789}, {19.773, 7.882}};
              PGpolygon polygon = {2, 1, pts};
              PQputf(param, "%polygon", &polygon);

       Getting a polygon value:

              PGpolygon polygon;
              if(PQgetf(result, tup_num, "%polygon", field_num, &polygon))
              {
                   // polygon.pts must be copied out if needed after clearing results
                   copy_points(polygon.npts, polygon.pts, ...);

                   PQclear(result);
                   // polygon.pts is now invalid
              }

INET & CIDR
       When  putting  an  inet  or cidr, all members must be set excluding the
       sa_len.

              typedef struct
              {
                   /* The address mask, 32 for a single IP. */
                   int mask;

                   /* When non-zero, the PGinet structure represents a cidr
                    * otherwise an inet.
                    */
                   int is_cidr;

                   /* the length in bytes of the sa_buf member. */
                   int sa_len;

                   /* the socket address buffer, contains the data.  This can
                    * be casted to a sockaddr, sockaddr_in, sockaddr_in6 or a
                    * sockaddr_storage structure. This buffer is 128 bytes so
                    * that it is large enough for a sockaddr_storage structure.
                    */
                   char sa_buf[128];
              } PGinet;

       Putting an inet or cidr:

              socklen_t len;
              PGinet inet;

              cli_fd = accept(srv_fd, (struct sockaddr *)inet.sa_buf, &len);
              if(cli_fd != -1)
              {
                   inet.is_cidr = 0;
                   inet.mask = 32;
                   PQputf(param, "%inet", &inet);
              }

       Getting an inet or cidr:

              PGinet inet;
              unsigned short port;

              /* gets an inet from field 2 and an int2 from field 6 */
              if(PQgetf(result, tup_num, "%inet %int2", 2, &inet, 6, &port))
              {
                   char ip[80];
                   struct sockaddr *sa = (struct sockaddr *)inet.sa_buf;

                   // converting a PGinet to an IPv4 or IPv6 address string
                   getnameinfo(sa, inet.sa_len, ip, sizeof(ip),
                        NULL, 0, NI_NUMERICHOST);

                   // The inet data type does not store a port.
                   if(sa->sa_family == AF_INET)
                        ((struct sockaddr_in *)sa)->sin_port = htons(port);
                   else
                        ((struct sockaddr_in6 *)sa)->sin6_port = htons(port);

                   printf("Connecting to %s:%d\n", ip, port);
                   connect(sock_fd, sa, inet.sa_len);
              }

MACADDR
       The PGmacaddr structure is used to put and get a macaddr.

              typedef struct
              {
                   int a;
                   int b;
                   int c;
                   int d;
                   int e;
                   int f;
              } PGmacaddr;

       Putting a macaddr value:

              PGmacaddr macaddr = {0, 1, 2, 3, 4, 5};
              PQputf(param, "%macaddr", &macaddr);

       Getting a macaddr value:

              PGmacaddr macaddr;
              PQgetf(result, tup_num, "%macaddr", field_num, &macaddr);

MONEY
       The money type is put/get as a PGmoney (64-bit  integer).   It  can  be
       converted to dollar and cents format by dividing by 100: double money =
       (double)money64 / 100.0;.  Pre 8.3 servers are limited to 32-bit  money
       values.

       Putting a money value:

              PGmoney money = 600000000054LL; // 6 billion dollars and 54 cents
              PQputf(param, "%money", money);

       Getting a money value:

              PQgetf(result, tup_num, "%money", field_num, &money);

BOOL
       The  bool  type is put/get as a PGbool.  To put true or false, use 1 or
       0.

       Putting a bool value:

              PGbool b = 1; // put true
              PQputf(param, "%bool", b);

       Getting a bool value:

              PGbool b;
              PQgetf(result, tup_num, "%bool", field_num, &b);

UUID
       The uuid type is put/get as a sequence of 16 bytes.  To put a  uuid  as
       text, use "%str".  NOTE: this type is not available on pre 8.3 servers.

       Putting a uuid value:

              PGuuid uuid = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15};
              PQputf(param, "%uuid", uuid);

       Getting a uuid value:

              PGuuid uuid;
              PQgetf(result, tup_num, "%uuid", field_num, &uuid);

       WARNING: The data provided on a put call is expected to be at least  16
       bytes.

OID
       Putting an oid value:

              Oid oid = 2318;
              PQputf(param, "%oid", oid);

       Getting an oid value:

              Oid oid;
              PQgetf(result, tup_num, "%oid", field_num, &oid);

EXAMPLES
       None.

AUTHOR
       A  contribution  of  eSilo, LLC. for the PostgreSQL Database Management
       System.  Written by Andrew Chernow and Merlin Moncure.

REPORTING BUGS
       Report bugs to <libpqtypes@esilo.com>.

COPYRIGHT
       Copyright (c) 2011 eSilo, LLC. All rights reserved.
       This is free software; see the source for copying conditions.  There is
       NO  warranty; not even for MERCHANTABILITY or  FITNESS FOR A PARTICULAR
       PURPOSE.

SEE ALSO
       PQgetf(3), PQputf(3).



libpqtypes                           2011                         pqt-specs(3)

libpqtypes home page