35.9. Functions Associated with the
COPY command in Postgres Pro has options to read from or write to the network connection used by libpq. The functions described in this section allow applications to take advantage of this capability by supplying or consuming copied data.
The overall process is that the application first issues the SQL
COPY command via
PQexec or one of the equivalent functions. The response to this (if there is no error in the command) will be a
PGresult object bearing a status code of
PGRES_COPY_IN (depending on the specified copy direction). The application should then use the functions of this section to receive or transmit data rows. When the data transfer is complete, another
PGresult object is returned to indicate success or failure of the transfer. Its status will be
PGRES_COMMAND_OK for success or
PGRES_FATAL_ERROR if some problem was encountered. At this point further SQL commands can be issued via
PQexec. (It is not possible to execute other SQL commands using the same connection while the
COPY operation is in progress.)
COPY command is issued via
PQexec in a string that could contain additional commands, the application must continue fetching results via
PQgetResult after completing the
COPY sequence. Only when
NULL is it certain that the
PQexec command string is done and it is safe to issue more commands.
PGresult object bearing one of these status values carries some additional data about the
COPY operation that is starting. This additional data is available using functions that are also used in connection with query results:
Returns the number of columns (fields) to be copied.
0 indicates the overall copy format is textual (rows separated by newlines, columns separated by separator characters, etc). 1 indicates the overall copy format is binary. See COPY for more information.
Returns the format code (0 for text, 1 for binary) associated with each column of the copy operation. The per-column format codes will always be zero when the overall copy format is textual, but the binary format can support both text and binary columns. (However, as of the current implementation of
COPY, only binary columns appear in a binary copy; so the per-column formats always match the overall format at present.)
These additional data values are only available when using protocol 3.0. When using protocol 2.0, all these functions will return 0.
35.9.1. Functions for Sending
These functions are used to send data during
COPY FROM STDIN. They will fail if called when the connection is not in
Sends data to the server during
int PQputCopyData(PGconn *conn, const char *buffer, int nbytes);
COPYdata in the specified
buffer, of length
nbytes, to the server. The result is 1 if the data was queued, zero if it was not queued because of full buffers (this will only happen in nonblocking mode), or -1 if an error occurred. (Use
PQerrorMessageto retrieve details if the return value is -1. If the value is zero, wait for write-ready and try again.)
The application can divide the
COPYdata stream into buffer loads of any convenient size. Buffer-load boundaries have no semantic significance when sending. The contents of the data stream must match the data format expected by the
COPYcommand; see COPY for details.
Sends end-of-data indication to the server during
int PQputCopyEnd(PGconn *conn, const char *errormsg);
COPY_INoperation successfully if
COPYis forced to fail, with the string pointed to by
errormsgused as the error message. (One should not assume that this exact error message will come back from the server, however, as the server might have already failed the
COPYfor its own reasons. Also note that the option to force failure does not work when using pre-3.0-protocol connections.)
The result is 1 if the termination message was sent; or in nonblocking mode, this may only indicate that the termination message was successfully queued. (In nonblocking mode, to be certain that the data has been sent, you should next wait for write-ready and call
PQflush, repeating until it returns zero.) Zero indicates that the function could not queue the termination message because of full buffers; this will only happen in nonblocking mode. (In this case, wait for write-ready and try the
PQputCopyEndcall again.) If a hard error occurs, -1 is returned; you can use
PQerrorMessageto retrieve details.
After successfully calling
PQgetResultto obtain the final result status of the
COPYcommand. One can wait for this result to be available in the usual way. Then return to normal operation.
35.9.2. Functions for Receiving
These functions are used to receive data during
COPY TO STDOUT. They will fail if called when the connection is not in
Receives data from the server during
int PQgetCopyData(PGconn *conn, char **buffer, int async);
Attempts to obtain another row of data from the server during a
COPY. Data is always returned one data row at a time; if only a partial row is available, it is not returned. Successful return of a data row involves allocating a chunk of memory to hold the data. The
bufferparameter must be non-
*bufferis set to point to the allocated memory, or to
NULLin cases where no buffer is returned. A non-
NULLresult buffer should be freed using
PQfreememwhen no longer needed.
When a row is successfully returned, the return value is the number of data bytes in the row (this will always be greater than zero). The returned string is always null-terminated, though this is probably only useful for textual
COPY. A result of zero indicates that the
COPYis still in progress, but no row is yet available (this is only possible when
asyncis true). A result of -1 indicates that the
COPYis done. A result of -2 indicates that an error occurred (consult
PQerrorMessagefor the reason).
asyncis true (not zero),
PQgetCopyDatawill not block waiting for input; it will return zero if the
COPYis still in progress but no complete row is available. (In this case wait for read-ready and then call
asyncis false (zero),
PQgetCopyDatawill block until data is available or the operation completes.
35.9.3. Obsolete Functions for
These functions represent older methods of handling
COPY. Although they still work, they are deprecated due to poor error handling, inconvenient methods of detecting end-of-data, and lack of support for binary or nonblocking transfers.
Reads a newline-terminated line of characters (transmitted by the server) into a buffer string of size
int PQgetline(PGconn *conn, char *buffer, int length);
This function copies up to
length-1 characters into the buffer and converts the terminating newline into a zero byte.
EOFat the end of input, 0 if the entire line has been read, and 1 if the buffer is full but the terminating newline has not yet been read.
Note that the application must check to see if a new line consists of the two characters
\., which indicates that the server has finished sending the results of the
COPYcommand. If the application might receive lines that are more than
length-1 characters long, care is needed to be sure it recognizes the
\.line correctly (and does not, for example, mistake the end of a long data line for a terminator line).
Reads a row of
COPYdata (transmitted by the server) into a buffer without blocking.
int PQgetlineAsync(PGconn *conn, char *buffer, int bufsize);
This function is similar to
PQgetline, but it can be used by applications that must read
COPYdata asynchronously, that is, without blocking. Having issued the
COPYcommand and gotten a
PGRES_COPY_OUTresponse, the application should call
PQgetlineAsyncuntil the end-of-data signal is detected.
PQgetline, this function takes responsibility for detecting end-of-data.
On each call,
PQgetlineAsyncwill return data if a complete data row is available in libpq's input buffer. Otherwise, no data is returned until the rest of the row arrives. The function returns -1 if the end-of-copy-data marker has been recognized, or 0 if no data is available, or a positive number giving the number of bytes of data returned. If -1 is returned, the caller must next call
PQendcopy, and then return to normal processing.
The data returned will not extend beyond a data-row boundary. If possible a whole row will be returned at one time. But if the buffer offered by the caller is too small to hold a row sent by the server, then a partial data row will be returned. With textual data this can be detected by testing whether the last returned byte is
\nor not. (In a binary
COPY, actual parsing of the
COPYdata format will be needed to make the equivalent determination.) The returned string is not null-terminated. (If you want to add a terminating null, be sure to pass a
bufsizeone smaller than the room actually available.)
Sends a null-terminated string to the server. Returns 0 if OK and
EOFif unable to send the string.
int PQputline(PGconn *conn, const char *string);
COPYdata stream sent by a series of calls to
PQputlinehas the same format as that returned by
PQgetlineAsync, except that applications are not obliged to send exactly one data row per
PQputlinecall; it is okay to send a partial line or multiple lines per call.
Before Postgres Pro protocol 3.0, it was necessary for the application to explicitly send the two characters
\.as a final line to indicate to the server that it had finished sending
COPYdata. While this still works, it is deprecated and the special meaning of
\.can be expected to be removed in a future release. It is sufficient to call
PQendcopyafter having sent the actual data.
Sends a non-null-terminated string to the server. Returns 0 if OK and
EOFif unable to send the string.
int PQputnbytes(PGconn *conn, const char *buffer, int nbytes);
This is exactly like
PQputline, except that the data buffer need not be null-terminated since the number of bytes to send is specified directly. Use this procedure when sending binary data.
Synchronizes with the server.
int PQendcopy(PGconn *conn);
This function waits until the server has finished the copying. It should either be issued when the last string has been sent to the server using
PQputlineor when the last string has been received from the server using
PQgetline. It must be issued or the server will get “out of sync” with the client. Upon return from this function, the server is ready to receive the next SQL command. The return value is 0 on successful completion, nonzero otherwise. (Use
PQerrorMessageto retrieve details if the return value is nonzero.)
PQgetResult, the application should respond to a
PGRES_COPY_OUTresult by executing
PQgetlinerepeatedly, followed by
PQendcopyafter the terminator line is seen. It should then return to the
PQgetResultreturns a null pointer. Similarly a
PGRES_COPY_INresult is processed by a series of
PQputlinecalls followed by
PQendcopy, then return to the
PQgetResultloop. This arrangement will ensure that a
COPYcommand embedded in a series of SQL commands will be executed correctly.
Older applications are likely to submit a
PQexecand assume that the transaction is done after
PQendcopy. This will work correctly only if the
COPYis the only SQL command in the command string.