Error Handling

33.8.1. Setting Callbacks

One simple method to catch errors and warnings is to set a specific action to be executed whenever a particular condition occurs. In general:

EXEC SQL WHENEVER condition action;

condition can be one of the following:

action can be one of the following:

The SQL standard only provides for the actions CONTINUE and GOTO (and GO TO ).

Here is an example that you might want to use in a simple program. It prints a simple message when a warning occurs and aborts the program when an error happens:

EXEC SQL WHENEVER SQLWARNING SQLPRINT;
EXEC SQL WHENEVER SQLERROR STOP;

The statement EXEC SQL WHENEVER is a directive of the SQL preprocessor, not a C statement. The error or warning actions that it sets apply to all embedded SQL statements that appear below the point where the handler is set, unless a different action was set for the same condition between the first EXEC SQL WHENEVER and the SQL statement causing the condition, regardless of the flow of control in the C program. So neither of the two following C program excerpts will have the desired effect:

/*
 * WRONG
 */
int main(int argc, char *argv[])
{
    ...
    if (verbose) {
        EXEC SQL WHENEVER SQLWARNING SQLPRINT;
    }
    ...
    EXEC SQL SELECT ...;
    ...
}

/*
 * WRONG
 */
int main(int argc, char *argv[])
{
    ...
    set_error_handler();
    ...
    EXEC SQL SELECT ...;
    ...
}

static void set_error_handler(void)
{
    EXEC SQL WHENEVER SQLERROR STOP;
}

33.8.2. sqlca

For more powerful error handling, the embedded SQL interface provides a global variable with the name sqlca (SQL communication area) that has the following structure:

struct
{
    char sqlcaid[8];
    long sqlabc;
    long sqlcode;
    struct
    {
        int sqlerrml;
        char sqlerrmc[SQLERRMC_LEN];
    } sqlerrm;
    char sqlerrp[8];
    long sqlerrd[6];
    char sqlwarn[8];
    char sqlstate[5];
} sqlca;

(In a multithreaded program, every thread automatically gets its own copy of sqlca . This works similarly to the handling of the standard C global variable errno .)

sqlca covers both warnings and errors. If multiple warnings or errors occur during the execution of a statement, then sqlca will only contain information about the last one.

If no error occurred in the last SQL statement, sqlca.sqlcode will be 0 and sqlca.sqlstate will be "00000" . If a warning or error occurred, then sqlca.sqlcode will be negative and sqlca.sqlstate will be different from "00000" . A positive sqlca.sqlcode indicates a harmless condition, such as that the last query returned zero rows. sqlcode and sqlstate are two different error code schemes; details appear below.

If the last SQL statement was successful, then sqlca.sqlerrd[1] contains the OID of the processed row, if applicable, and sqlca.sqlerrd[2] contains the number of processed or returned rows, if applicable to the command.

In case of an error or warning, sqlca.sqlerrm.sqlerrmc will contain a string that describes the error. The field sqlca.sqlerrm.sqlerrml contains the length of the error message that is stored in sqlca.sqlerrm.sqlerrmc (the result of strlen() , not really interesting for a C programmer). Note that some messages are too long to fit in the fixed-size sqlerrmc array; they will be truncated.

In case of a warning, sqlca.sqlwarn[2] is set to W . (In all other cases, it is set to something different from W .) If sqlca.sqlwarn[1] is set to W , then a value was truncated when it was stored in a host variable. sqlca.sqlwarn[0] is set to W if any of the other elements are set to indicate a warning.

The fields sqlcaid , sqlcabc , sqlerrp , and the remaining elements of sqlerrd and sqlwarn currently contain no useful information.

The structure sqlca is not defined in the SQL standard, but is implemented in several other SQL database systems. The definitions are similar at the core, but if you want to write portable applications, then you should investigate the different implementations carefully.

Here is one example that combines the use of WHENEVER and sqlca , printing out the contents of sqlca when an error occurs. This is perhaps useful for debugging or prototyping applications, before installing a more "user-friendly" error handler.

EXEC SQL WHENEVER SQLERROR CALL print_sqlca();

void
print_sqlca()
{
    fprintf(stderr, "==== sqlca ====\n");
    fprintf(stderr, "sqlcode: %ld\n", sqlca.sqlcode);
    fprintf(stderr, "sqlerrm.sqlerrml: %d\n", sqlca.sqlerrm.sqlerrml);
    fprintf(stderr, "sqlerrm.sqlerrmc: %s\n", sqlca.sqlerrm.sqlerrmc);
    fprintf(stderr, "sqlerrd: %ld %ld %ld %ld %ld %ld\n", sqlca.sqlerrd[0],sqlca.sqlerrd[1],sqlca.sqlerrd[2],
                                                          sqlca.sqlerrd[3],sqlca.sqlerrd[4],sqlca.sqlerrd[5]);
    fprintf(stderr, "sqlwarn: %d %d %d %d %d %d %d %d\n", sqlca.sqlwarn[0], sqlca.sqlwarn[1], sqlca.sqlwarn[2],
                                                          sqlca.sqlwarn[3], sqlca.sqlwarn[4], sqlca.sqlwarn[5],
                                                          sqlca.sqlwarn[6], sqlca.sqlwarn[7]);
    fprintf(stderr, "sqlstate: %5s\n", sqlca.sqlstate);
    fprintf(stderr, "===============\n");
}

The result could look as follows (here an error due to a misspelled table name):

==== sqlca ====
sqlcode: -400
sqlerrm.sqlerrml: 49
sqlerrm.sqlerrmc: relation "pg_databasep" does not exist on line 38
sqlerrd: 0 0 0 0 0 0
sqlwarn: 0 0 0 0 0 0 0 0
sqlstate: 42P01
===============

33.8.3. SQLSTATE vs. SQLCODE

The fields sqlca.sqlstate and sqlca.sqlcode are two different schemes that provide error codes. Both are derived from the SQL standard, but SQLCODE has been marked deprecated in the SQL-92 edition of the standard and has been dropped in later editions. Therefore, new applications are strongly encouraged to use SQLSTATE .

SQLSTATE is a five-character array. The five characters contain digits or upper-case letters that represent codes of various error and warning conditions. SQLSTATE has a hierarchical scheme: the first two characters indicate the general class of the condition, the last three characters indicate a subclass of the general condition. A successful state is indicated by the code 00000 . The SQLSTATE codes are for the most part defined in the SQL standard. The PostgreSQL server natively supports SQLSTATE error codes; therefore a high degree of consistency can be achieved by using this error code scheme throughout all applications. For further information see Appendix A .

SQLCODE , the deprecated error code scheme, is a simple integer. A value of 0 indicates success, a positive value indicates success with additional information, a negative value indicates an error. The SQL standard only defines the positive value +100, which indicates that the last command returned or affected zero rows, and no specific negative values. Therefore, this scheme can only achieve poor portability and does not have a hierarchical code assignment. Historically, the embedded SQL processor for PostgreSQL has assigned some specific SQLCODE values for its use, which are listed below with their numeric value and their symbolic name. Remember that these are not portable to other SQL implementations. To simplify the porting of applications to the SQLSTATE scheme, the corresponding SQLSTATE is also listed. There is, however, no one-to-one or one-to-many mapping between the two schemes (indeed it is many-to-many), so you should consult the global SQLSTATE listing in Appendix A in each case.

These are the assigned SQLCODE values: