Merge branch 'devel'

NEWS

@@ -1,3 +1,18 @@
+What's new in psycopg 2.4.1
+---------------------------
+
+  - Use own parser for bytea output, not requiring anymore the libpq 9.0
+    to parse the hex format.
+  - Don't fail connection if the client encoding is a non-normalized
+    variant. Issue reported by Peter Eisentraut.
+  - Correctly detect an empty query sent to the backend (ticket #46).
+  - Fixed a SystemError clobbering libpq errors raised without SQLSTATE.
+    Bug vivisectioned by Eric Snow.
+  - Fixed interaction between NamedTuple and server-side cursors.
+  - Allow to specify --static-libpq on setup.py command line instead of
+    just in 'setup.cfg'. Patch provided by Matthew Ryan (ticket #48).
+
+
 What's new in psycopg 2.4
 -------------------------
 
@@ -163,6 +178,8 @@ doc/html/advanced.html.
   - Fixed TimestampFromTicks() and TimeFromTicks() for seconds >= 59.5.
   - Fixed spurious exception raised when calling C typecasters from Python
     ones.
+
+
 What's new in psycopg 2.0.14
 ----------------------------
 

ZPsycopgDA/DA.py

@@ -16,7 +16,7 @@
 # their work without bothering about the module dependencies.
 
 
-ALLOWED_PSYCOPG_VERSIONS = ('2.4-beta1', '2.4-beta2', '2.4')
+ALLOWED_PSYCOPG_VERSIONS = ('2.4-beta1', '2.4-beta2', '2.4', '2.4.1')
 
 import sys
 import time

doc/src/faq.rst

@@ -97,7 +97,9 @@ Psycopg converts :sql:`decimal`\/\ :sql:`numeric` database types into Python `!D
 Transferring binary data from PostgreSQL 9.0 doesn't work.
     PostgreSQL 9.0 uses by default `the "hex" format`__ to transfer
     :sql:`bytea` data: the format can't be parsed by the libpq 8.4 and
-    earlier. Three options to solve the problem are:
+    earlier. The problem is solved in Psycopg 2.4.1, that uses its own parser
+    for the :sql:`bytea` format. For previous Psycopg releases, three options
+    to solve the problem are:
 
     - set the bytea_output__ parameter to ``escape`` in the server;
     - execute the database command ``SET bytea_output TO escape;`` in the

doc/src/usage.rst

@@ -271,6 +271,10 @@ the SQL string that would be sent to the database.
   .. versionchanged:: 2.4
      only strings were supported before.
 
+  .. versionchanged:: 2.4.1
+     can parse the 'hex' format from 9.0 servers without relying on the
+     version of the client library.
+
   .. note::
 
     In Python 2, if you have binary data in a `!str` object, you can pass them
@@ -282,18 +286,15 @@ the SQL string that would be sent to the database.
 
   .. warning::
 
-     PostgreSQL 9 uses by default `a new "hex" format`__ to emit :sql:`bytea`
+     Since version 9.0 PostgreSQL uses by default `a new "hex" format`__ to
-     fields. Unfortunately this format can't be parsed by libpq versions
+     emit :sql:`bytea` fields. Starting from Psycopg 2.4.1 the format is
-     before 9.0. This means that using a library client with version lesser
+     correctly supported.  If you use a previous version you will need some
-     than 9.0 to talk with a server 9.0 or later you may have problems
+     extra care when receiving bytea from PostgreSQL: you must have at least
-     receiving :sql:`bytea` data. To work around this problem you can set the
+     the libpq 9.0 installed on the client or alternatively you can set the
-     `bytea_output`__ parameter to ``escape``, either in the server
+     `bytea_output`__ configutation parameter to ``escape``, either in the
-     configuration or in the client session using a query such as ``SET
+     server configuration file or in the client session (using a query such as
-     bytea_output TO escape;`` before trying to receive binary data.
+     ``SET bytea_output TO escape;``) before receiving binary data.
      
-     Starting from Psycopg 2.4 this condition is detected and signaled with a
-     `~psycopg2.InterfaceError`.
-
      .. __: http://www.postgresql.org/docs/9.0/static/datatype-binary.html
      .. __: http://www.postgresql.org/docs/9.0/static/runtime-config-client.html#GUC-BYTEA-OUTPUT
 

lib/extras.py

@@ -28,7 +28,6 @@ and classes untill a better place in the distribution is found.
 import os
 import sys
 import time
-import codecs
 import warnings
 import re as regex
 
@@ -291,21 +290,28 @@ class NamedTupleCursor(_cursor):
             return nt(*t)
 
     def fetchmany(self, size=None):
+        ts = _cursor.fetchmany(self, size)
         nt = self.Record
         if nt is None:
             nt = self.Record = self._make_nt()
-        ts = _cursor.fetchmany(self, size)
         return [nt(*t) for t in ts]
 
     def fetchall(self):
+        ts = _cursor.fetchall(self)
         nt = self.Record
         if nt is None:
             nt = self.Record = self._make_nt()
-        ts = _cursor.fetchall(self)
         return [nt(*t) for t in ts]
 
     def __iter__(self):
-        return iter(self.fetchall())
+        # Invoking _cursor.__iter__(self) goes to infinite recursion,
+        # so we do pagination by hand
+        while 1:
+            recs = self.fetchmany(self.itersize)
+            if not recs:
+                return
+            for rec in recs:
+                yield rec
 
     try:
         from collections import namedtuple

psycopg/connection_int.c

@@ -236,10 +236,45 @@ conn_get_standard_conforming_strings(PGconn *pgconn)
     return equote;
 }
 
+
+/* Remove irrelevant chars from encoding name and turn it uppercase.
+ *
+ * Return a buffer allocated on Python heap,
+ * NULL and set an exception on error.
+ */
+static char *
+clean_encoding_name(const char *enc)
+{
+    const char *i = enc;
+    char *rv, *j;
+
+    /* convert to upper case and remove '-' and '_' from string */
+    if (!(j = rv = PyMem_Malloc(strlen(enc) + 1))) {
+        PyErr_NoMemory();
+        return NULL;
+    }
+
+    while (*i) {
+        if (!isalnum(*i)) {
+            ++i;
+        }
+        else {
+            *j++ = toupper(*i++);
+        }
+    }
+    *j = '\0';
+
+    Dprintf("clean_encoding_name: %s -> %s", enc, rv);
+
+    return rv;
+}
+
 /* Convert a PostgreSQL encoding to a Python codec.
  *
  * Return a new copy of the codec name allocated on the Python heap,
  * NULL with exception in case of error.
+ *
+ * 'enc' should be already normalized (uppercase, no - or _).
  */
 static char *
 conn_encoding_to_codec(const char *enc)
@@ -285,7 +320,7 @@ exit:
 static int
 conn_read_encoding(connectionObject *self, PGconn *pgconn)
 {
-    char *enc = NULL, *codec = NULL, *j;
+    char *enc = NULL, *codec = NULL;
     const char *tmp;
     int rv = -1;
 
@@ -297,16 +332,10 @@ conn_read_encoding(connectionObject *self, PGconn *pgconn)
         goto exit;
     }
 
-    if (!(enc = PyMem_Malloc(strlen(tmp)+1))) {
+    if (!(enc = clean_encoding_name(tmp))) {
-        PyErr_NoMemory();
         goto exit;
     }
 
-    /* turn encoding in uppercase */
-    j = enc;
-    while (*tmp) { *j++ = toupper(*tmp++); }
-    *j = '\0';
-
     /* Look for this encoding in Python codecs. */
     if (!(codec = conn_encoding_to_codec(enc))) {
         goto exit;
@@ -965,21 +994,23 @@ conn_set_client_encoding(connectionObject *self, const char *enc)
     PGresult *pgres = NULL;
     char *error = NULL;
     char query[48];
-    int res = 0;
+    int res = 1;
-    char *codec;
+    char *codec = NULL;
+    char *clean_enc = NULL;
 
     /* If the current encoding is equal to the requested one we don't
        issue any query to the backend */
     if (strcmp(self->encoding, enc) == 0) return 0;
 
     /* We must know what python codec this encoding is. */
-    if (!(codec = conn_encoding_to_codec(enc))) { return -1; }
+    if (!(clean_enc = clean_encoding_name(enc))) { goto exit; }
+    if (!(codec = conn_encoding_to_codec(clean_enc))) { goto exit; }
 
     Py_BEGIN_ALLOW_THREADS;
     pthread_mutex_lock(&self->lock);
 
     /* set encoding, no encoding string is longer than 24 bytes */
-    PyOS_snprintf(query, 47, "SET client_encoding = '%s'", enc);
+    PyOS_snprintf(query, 47, "SET client_encoding = '%s'", clean_enc);
 
     /* abort the current transaction, to set the encoding ouside of
        transactions */
@@ -994,21 +1025,18 @@ conn_set_client_encoding(connectionObject *self, const char *enc)
     /* no error, we can proceeed and store the new encoding */
     {
         char *tmp = self->encoding;
-        self->encoding = NULL;
+        self->encoding = clean_enc;
         PyMem_Free(tmp);
-    }
+        clean_enc = NULL;
-    if (!(self->encoding = psycopg_strdup(enc, 0))) {
-        res = 1;  /* don't call pq_complete_error below */
-        goto endlock;
     }
 
     /* Store the python codec too. */
     {
         char *tmp = self->codec;
-        self->codec = NULL;
+        self->codec = codec;
         PyMem_Free(tmp);
+        codec = NULL;
     }
-    self->codec = codec;
 
     Dprintf("conn_set_client_encoding: set encoding to %s (codec: %s)",
             self->encoding, self->codec);
@@ -1021,6 +1049,10 @@ endlock:
     if (res < 0)
         pq_complete_error(self, &pgres, &error);
 
+exit:
+    PyMem_Free(clean_enc);
+    PyMem_Free(codec);
+
     return res;
 }
 

psycopg/connection_type.c

@@ -423,36 +423,18 @@ static PyObject *
 psyco_conn_set_client_encoding(connectionObject *self, PyObject *args)
 {
     const char *enc;
-    char *buffer, *dest;
     PyObject *rv = NULL;
-    Py_ssize_t len;
 
     EXC_IF_CONN_CLOSED(self);
     EXC_IF_CONN_ASYNC(self, set_client_encoding);
     EXC_IF_TPC_PREPARED(self, set_client_encoding);
 
-    if (!PyArg_ParseTuple(args, "s#", &enc, &len)) return NULL;
+    if (!PyArg_ParseTuple(args, "s", &enc)) return NULL;
 
-    /* convert to upper case and remove '-' and '_' from string */
+    if (conn_set_client_encoding(self, enc) == 0) {
-    if (!(dest = buffer = PyMem_Malloc(len+1))) {
-        return PyErr_NoMemory();
-    }
-
-    while (*enc) {
-        if (*enc == '_' || *enc == '-') {
-            ++enc;
-        }
-        else {
-            *dest++ = toupper(*enc++);
-        }
-    }
-    *dest = '\0';
-
-    if (conn_set_client_encoding(self, buffer) == 0) {
         Py_INCREF(Py_None);
         rv = Py_None;
     }
-    PyMem_Free(buffer);
     return rv;
 }
 

psycopg/pqpath.c

@@ -172,16 +172,19 @@ pq_raise(connectionObject *conn, cursorObject *curs, PGresult *pgres)
     if (pgres) {
         err = PQresultErrorMessage(pgres);
         if (err != NULL) {
+            Dprintf("pq_raise: PQresultErrorMessage: err=%s", err);
             code = PQresultErrorField(pgres, PG_DIAG_SQLSTATE);
         }
     }
-    if (err == NULL)
+    if (err == NULL) {
         err = PQerrorMessage(conn->pgconn);
+        Dprintf("pq_raise: PQerrorMessage: err=%s", err);
+    }
 
     /* if the is no error message we probably called pq_raise without reason:
        we need to set an exception anyway because the caller will probably
        raise and a meaningful message is better than an empty one */
-    if (err == NULL) {
+    if (err == NULL || err[0] == '\0') {
         PyErr_SetString(Error, "psycopg went psycotic without error set");
         return;
     }
@@ -191,9 +194,15 @@ pq_raise(connectionObject *conn, cursorObject *curs, PGresult *pgres)
     if (code != NULL) {
         exc = exception_from_sqlstate(code);
     }
+    else {
+        /* Fallback if there is no exception code (reported happening e.g.
+         * when the connection is closed). */
+        exc = DatabaseError;
+    }
 
     /* try to remove the initial "ERROR: " part from the postgresql error */
     err2 = strip_severity(err);
+    Dprintf("pq_raise: err2=%s", err2);
 
     psyco_set_error(exc, curs, err2, err, code);
 }
@@ -1355,6 +1364,13 @@ pq_fetch(cursorObject *curs)
         /* don't clear curs->pgres, because it contains the results! */
         break;
 
+    case PGRES_EMPTY_QUERY:
+        PyErr_SetString(ProgrammingError,
+            "can't execute an empty query");
+        IFCLEARPGRES(curs->pgres);
+        ex = -1;
+        break;
+
     default:
         Dprintf("pq_fetch: uh-oh, something FAILED: pgconn = %p", curs->conn);
         pq_raise(curs->conn, curs, NULL);

psycopg/typecast_binary.c

@@ -40,7 +40,7 @@ chunk_dealloc(chunkObject *self)
         FORMAT_CODE_PY_SSIZE_T,
         self->base, self->len
       );
-    PQfreemem(self->base);
+    PyMem_Free(self->base);
     Py_TYPE(self)->tp_free((PyObject *)self);
 }
 
@@ -127,95 +127,185 @@ PyTypeObject chunkType = {
     chunk_doc                   /* tp_doc */
 };
 
-static PyObject *
+
+static char *psycopg_parse_hex(
+        const char *bufin, Py_ssize_t sizein, Py_ssize_t *sizeout);
+static char *psycopg_parse_escape(
+        const char *bufin, Py_ssize_t sizein, Py_ssize_t *sizeout);
+
+/* The function is not static and not hidden as we use ctypes to test it. */
+PyObject *
 typecast_BINARY_cast(const char *s, Py_ssize_t l, PyObject *curs)
 {
     chunkObject *chunk = NULL;
     PyObject *res = NULL;
-    char *str = NULL, *buffer = NULL;
+    char *buffer = NULL;
-    size_t len;
+    Py_ssize_t len;
 
     if (s == NULL) {Py_INCREF(Py_None); return Py_None;}
 
-    /* PQunescapeBytea absolutely wants a 0-terminated string and we don't
+    if (s[0] == '\\' && s[1] == 'x') {
-       want to copy the whole buffer, right? Wrong, but there isn't any other
+        /* This is a buffer escaped in hex format: libpq before 9.0 can't
-       way <g> */
+         * parse it and we can't detect reliably the libpq version at runtime.
-    if (s[l] != '\0') {
+         * So the only robust option is to parse it ourselves - luckily it's
-        if ((buffer = PyMem_Malloc(l+1)) == NULL) {
+         * an easy format.
-            PyErr_NoMemory();
+         */
-            goto fail;
+        if (NULL == (buffer = psycopg_parse_hex(s, l, &len))) {
+            goto exit;
         }
-        /* Py_ssize_t->size_t cast is safe, as long as the Py_ssize_t is
-         * >= 0: */
-        assert (l >= 0);
-        strncpy(buffer, s, (size_t) l);
-
-        buffer[l] = '\0';
-        s = buffer;
     }
-    str = (char*)PQunescapeBytea((unsigned char*)s, &len);
+    else {
-    Dprintf("typecast_BINARY_cast: unescaped " FORMAT_CODE_SIZE_T " bytes",
+        /* This is a buffer in the classic bytea format. So we can handle it
-      len);
+         * to the PQunescapeBytea to have it parsed, rignt? ...Wrong. We
-
+         * could, but then we'd have to record whether buffer was allocated by
-    /* The type of the second parameter to PQunescapeBytea is size_t *, so it's
+         * Python or by the libpq to dispose it properly. Furthermore the
-     * possible (especially with Python < 2.5) to get a return value too large
+         * PQunescapeBytea interface is not the most brilliant as it wants a
-     * to fit into a Python container. */
+         * null-terminated string even if we have known its length thus
-    if (len > (size_t) PY_SSIZE_T_MAX) {
+         * requiring a useless memcpy and strlen.
-      PyErr_SetString(PyExc_IndexError, "PG buffer too large to fit in Python"
+         * So we'll just have our better integrated parser, let's finish this
-                                        " buffer.");
+         * story.
-      goto fail;
+         */
-    }
+        if (NULL == (buffer = psycopg_parse_escape(s, l, &len))) {
-
+            goto exit;
-    /* Check the escaping was successful */
+        }
-    if (s[0] == '\\' && s[1] == 'x'     /* input encoded in hex format */
-        && str[0] == 'x'                /* output resulted in an 'x' */
-        && s[2] != '7' && s[3] != '8')  /* input wasn't really an x (0x78) */
-    {
-        PyErr_SetString(InterfaceError,
-            "can't receive bytea data from server >= 9.0 with the current "
-            "libpq client library: please update the libpq to at least 9.0 "
-            "or set bytea_output to 'escape' in the server config "
-            "or with a query");
-        goto fail;
     }
 
     chunk = (chunkObject *) PyObject_New(chunkObject, &chunkType);
-    if (chunk == NULL) goto fail;
+    if (chunk == NULL) goto exit;
 
-    /* **Transfer** ownership of str's memory to the chunkObject: */
+    /* **Transfer** ownership of buffer's memory to the chunkObject: */
-    chunk->base = str;
+    chunk->base = buffer;
-    str = NULL;
+    buffer = NULL;
+    chunk->len = (Py_ssize_t)len;
 
-    /* size_t->Py_ssize_t cast was validated above: */
-    chunk->len = (Py_ssize_t) len;
 #if PY_MAJOR_VERSION < 3
     if ((res = PyBuffer_FromObject((PyObject *)chunk, 0, chunk->len)) == NULL)
-        goto fail;
+        goto exit;
 #else
     if ((res = PyMemoryView_FromObject((PyObject*)chunk)) == NULL)
-        goto fail;
+        goto exit;
 #endif
-    /* PyBuffer_FromObject() created a new reference.  We'll release our
-     * reference held in 'chunk' in the 'cleanup' clause. */
 
-    goto cleanup;
+exit:
-    fail:
+    Py_XDECREF((PyObject *)chunk);
-      assert (PyErr_Occurred());
+    PyMem_Free(buffer);
-      if (res != NULL) {
-          Py_DECREF(res);
-          res = NULL;
-      }
-      /* Fall through to cleanup: */
-    cleanup:
-      if (chunk != NULL) {
-          Py_DECREF((PyObject *) chunk);
-      }
-      if (str != NULL) {
-          /* str's mem was allocated by PQunescapeBytea; must use PQfreemem: */
-          PQfreemem(str);
-      }
-      /* We allocated buffer with PyMem_Malloc; must use PyMem_Free: */
-      PyMem_Free(buffer);
 
-      return res;
+    return res;
 }
+
+
+static const char hex_lut[128] = {
+    -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+    -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+    -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+     0,  1,  2,  3,  4,  5,  6,  7,  8,  9, -1, -1, -1, -1, -1, -1,
+    -1, 10, 11, 12, 13, 14, 15, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+    -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+    -1, 10, 11, 12, 13, 14, 15, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+    -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
+};
+
+/* Parse a bytea output buffer encoded in 'hex' format.
+ *
+ * the format is described in
+ * http://www.postgresql.org/docs/9.0/static/datatype-binary.html
+ *
+ * Parse the buffer in 'bufin', whose length is 'sizein'.
+ * Return a new buffer allocated by PyMem_Malloc and set 'sizeout' to its size.
+ * In case of error set an exception and return NULL.
+ */
+static char *
+psycopg_parse_hex(const char *bufin, Py_ssize_t sizein, Py_ssize_t *sizeout)
+{
+    char *ret = NULL;
+    const char *bufend = bufin + sizein;
+    const char *pi = bufin + 2;     /* past the \x */
+    char *bufout;
+    char *po;
+
+    po = bufout = PyMem_Malloc((sizein - 2) >> 1);   /* output size upper bound */
+    if (NULL == bufout) {
+        PyErr_NoMemory();
+        goto exit;
+    }
+
+    /* Implementation note: we call this function upon database response, not
+     * user input (because we are parsing the output format of a buffer) so we
+     * don't expect errors. On bad input we reserve the right to return a bad
+     * output, not an error.
+     */
+    while (pi < bufend) {
+        char c;
+        while (-1 == (c = hex_lut[*pi++ & '\x7f'])) {
+            if (pi >= bufend) { goto endloop; }
+        }
+        *po = c << 4;
+
+        while (-1 == (c = hex_lut[*pi++ & '\x7f'])) {
+            if (pi >= bufend) { goto endloop; }
+        }
+        *po++ |= c;
+    }
+endloop:
+
+    ret = bufout;
+    *sizeout = po - bufout;
+
+exit:
+    return ret;
+}
+
+/* Parse a bytea output buffer encoded in 'escape' format.
+ *
+ * the format is described in
+ * http://www.postgresql.org/docs/9.0/static/datatype-binary.html
+ *
+ * Parse the buffer in 'bufin', whose length is 'sizein'.
+ * Return a new buffer allocated by PyMem_Malloc and set 'sizeout' to its size.
+ * In case of error set an exception and return NULL.
+ */
+static char *
+psycopg_parse_escape(const char *bufin, Py_ssize_t sizein, Py_ssize_t *sizeout)
+{
+    char *ret = NULL;
+    const char *bufend = bufin + sizein;
+    const char *pi = bufin;
+    char *bufout;
+    char *po;
+
+    po = bufout = PyMem_Malloc(sizein);   /* output size upper bound */
+    if (NULL == bufout) {
+        PyErr_NoMemory();
+        goto exit;
+    }
+
+    while (pi < bufend) {
+        if (*pi != '\\') {
+            /* Unescaped char */
+            *po++ = *pi++;
+            continue;
+        }
+        if ((pi[1] >= '0' && pi[1] <= '3') &&
+            (pi[2] >= '0' && pi[2] <= '7') &&
+            (pi[3] >= '0' && pi[3] <= '7'))
+        {
+            /* Escaped octal value */
+            *po++ = ((pi[1] - '0') << 6) |
+                    ((pi[2] - '0') << 3) |
+                    ((pi[3] - '0'));
+            pi += 4;
+        }
+        else {
+            /* Escaped char */
+            *po++ = pi[1];
+            pi += 2;
+        }
+    }
+
+    ret = bufout;
+    *sizeout = po - bufout;
+
+exit:
+    return ret;
+}
+

setup.py

@@ -79,7 +79,7 @@ except ImportError:
 # Take a look at http://www.python.org/dev/peps/pep-0386/
 # for a consistent versioning pattern.
 
-PSYCOPG_VERSION = '2.4'
+PSYCOPG_VERSION = '2.4.1'
 
 version_flags   = ['dt', 'dec']
 
@@ -133,6 +133,7 @@ class psycopg_build_ext(build_ext):
         self.mx_include_dir = None
         self.use_pydatetime = 1
         self.have_ssl = have_ssl
+        self.static_libpq = static_libpq
         self.pg_config = None
 
     def get_compiler(self):
@@ -263,7 +264,7 @@ or with the pg_config option in 'setup.cfg'.
             sys.exit(1)
 
         self.include_dirs.append(".")
-        if static_libpq:
+        if self.static_libpq:
             if not self.link_objects: self.link_objects = []
             self.link_objects.append(
                     os.path.join(self.get_pg_config("libdir"), "libpq.a"))

tests/__init__.py

@@ -27,17 +27,6 @@ import sys
 from testconfig import dsn
 from testutils import unittest
 
-# If connection to test db fails, bail out early.
-import psycopg2
-try:
-    cnn = psycopg2.connect(dsn)
-except Exception, e:
-    print "Failed connection to test db:", e.__class__.__name__, e
-    print "Please set env vars 'PSYCOPG2_TESTDB*' to valid values."
-    sys.exit(1)
-else:
-    cnn.close()
-
 import bug_gc
 import bugX000
 import extras_dictcursor
@@ -57,6 +46,17 @@ import test_green
 import test_cancel
 
 def test_suite():
+    # If connection to test db fails, bail out early.
+    import psycopg2
+    try:
+        cnn = psycopg2.connect(dsn)
+    except Exception, e:
+        print "Failed connection to test db:", e.__class__.__name__, e
+        print "Please set env vars 'PSYCOPG2_TESTDB*' to valid values."
+        sys.exit(1)
+    else:
+        cnn.close()
+
     suite = unittest.TestSuite()
     suite.addTest(bug_gc.test_suite())
     suite.addTest(bugX000.test_suite())

tests/extras_dictcursor.py

@@ -14,9 +14,11 @@
 # FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
 # License for more details.
 
+import time
+from datetime import timedelta
 import psycopg2
 import psycopg2.extras
-from testutils import unittest, skip_if_no_namedtuple
+from testutils import unittest, skip_before_postgres, skip_if_no_namedtuple
 from testconfig import dsn
 
 
@@ -261,6 +263,53 @@ class NamedTupleCursorTest(unittest.TestCase):
         finally:
             NamedTupleCursor._make_nt = f_orig
 
+    @skip_if_no_namedtuple
+    @skip_before_postgres(8, 0)
+    def test_named(self):
+        curs = self.conn.cursor('tmp')
+        curs.execute("""select i from generate_series(0,9) i""")
+        recs = []
+        recs.extend(curs.fetchmany(5))
+        recs.append(curs.fetchone())
+        recs.extend(curs.fetchall())
+        self.assertEqual(range(10), [t.i for t in recs])
+
+    @skip_if_no_namedtuple
+    def test_named_fetchone(self):
+        curs = self.conn.cursor('tmp')
+        curs.execute("""select 42 as i""")
+        t = curs.fetchone()
+        self.assertEqual(t.i, 42)
+
+    @skip_if_no_namedtuple
+    def test_named_fetchmany(self):
+        curs = self.conn.cursor('tmp')
+        curs.execute("""select 42 as i""")
+        recs = curs.fetchmany(10)
+        self.assertEqual(recs[0].i, 42)
+
+    @skip_if_no_namedtuple
+    def test_named_fetchall(self):
+        curs = self.conn.cursor('tmp')
+        curs.execute("""select 42 as i""")
+        recs = curs.fetchall()
+        self.assertEqual(recs[0].i, 42)
+
+    @skip_if_no_namedtuple
+    @skip_before_postgres(8, 2)
+    def test_not_greedy(self):
+        curs = self.conn.cursor('tmp')
+        curs.itersize = 2
+        curs.execute("""select clock_timestamp() as ts from generate_series(1,3)""")
+        recs = []
+        for t in curs:
+            time.sleep(0.01)
+            recs.append(t)
+
+        # check that the dataset was not fetched in a single gulp
+        self.assert_(recs[1].ts - recs[0].ts < timedelta(seconds=0.005))
+        self.assert_(recs[2].ts - recs[1].ts > timedelta(seconds=0.0099))
+
 
 def test_suite():
     return unittest.TestLoader().loadTestsFromName(__name__)

tests/test_connection.py

@@ -22,6 +22,7 @@
 # FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
 # License for more details.
 
+import os
 import time
 import threading
 from testutils import unittest, decorate_all_tests, skip_before_postgres
@@ -141,6 +142,19 @@ class ConnectionTests(unittest.TestCase):
         cur.execute("select 'foo'::text;")
         self.assertEqual(cur.fetchone()[0], u'foo')
 
+    def test_connect_nonnormal_envvar(self):
+        # We must perform encoding normalization at connection time
+        self.conn.close()
+        oldenc = os.environ.get('PGCLIENTENCODING')
+        os.environ['PGCLIENTENCODING'] = 'utf-8'    # malformed spelling
+        try:
+            self.conn = psycopg2.connect(dsn)
+        finally:
+            if oldenc is not None:
+                os.environ['PGCLIENTENCODING'] = oldenc
+            else:
+                del os.environ['PGCLIENTENCODING']
+
     def test_weakref(self):
         from weakref import ref
         conn = psycopg2.connect(dsn)

tests/test_cursor.py

@@ -37,6 +37,12 @@ class CursorTests(unittest.TestCase):
     def tearDown(self):
         self.conn.close()
 
+    def test_empty_query(self):
+        cur = self.conn.cursor()
+        self.assertRaises(psycopg2.ProgrammingError, cur.execute, "")
+        self.assertRaises(psycopg2.ProgrammingError, cur.execute, " ")
+        self.assertRaises(psycopg2.ProgrammingError, cur.execute, ";")
+
     def test_executemany_propagate_exceptions(self):
         conn = self.conn
         cur = conn.cursor()

tests/testutils.py

@@ -140,24 +140,6 @@ def skip_if_no_namedtuple(f):
     return skip_if_no_namedtuple_
 
 
-def skip_if_broken_hex_binary(f):
-    """Decorator to detect libpq < 9.0 unable to parse bytea in hex format"""
-    def cope_with_hex_binary_(self):
-        from psycopg2 import InterfaceError
-        try:
-            return f(self)
-        except InterfaceError, e:
-            if '9.0' in str(e) and self.conn.server_version >= 90000:
-                return self.skipTest(
-                    # FIXME: we are only assuming the libpq is older here,
-                    # but we don't have a reliable way to detect the libpq
-                    # version, not pre-9 at least.
-                    "bytea broken with server >= 9.0, libpq < 9")
-            else:
-                raise
-
-    return cope_with_hex_binary_
-
 def skip_if_no_iobase(f):
     """Skip a test if io.TextIOBase is not available."""
     def skip_if_no_iobase_(self):

tests/types_basic.py

@@ -28,7 +28,7 @@ except:
     pass
 import sys
 import testutils
-from testutils import unittest, skip_if_broken_hex_binary
+from testutils import unittest, decorate_all_tests
 from testconfig import dsn
 
 import psycopg2
@@ -116,7 +116,6 @@ class TypesBasicTests(unittest.TestCase):
         s = self.execute("SELECT %s AS foo", (float("-inf"),))
         self.failUnless(str(s) == "-inf", "wrong float quoting: " + str(s))      
 
-    @skip_if_broken_hex_binary
     def testBinary(self):
         if sys.version_info[0] < 3:
             s = ''.join([chr(x) for x in range(256)])
@@ -143,7 +142,6 @@ class TypesBasicTests(unittest.TestCase):
             b = psycopg2.Binary(bytes([]))
             self.assertEqual(str(b), "''::bytea")
 
-    @skip_if_broken_hex_binary
     def testBinaryRoundTrip(self):
         # test to make sure buffers returned by psycopg2 are
         # understood by execute:
@@ -191,7 +189,6 @@ class TypesBasicTests(unittest.TestCase):
         s = self.execute("SELECT '{}'::text AS foo")
         self.failUnlessEqual(s, "{}")
 
-    @skip_if_broken_hex_binary
     @testutils.skip_from_python(3)
     def testTypeRoundtripBuffer(self):
         o1 = buffer("".join(map(chr, range(256))))
@@ -204,7 +201,6 @@ class TypesBasicTests(unittest.TestCase):
         self.assertEqual(type(o1), type(o2))
         self.assertEqual(str(o1), str(o2))
 
-    @skip_if_broken_hex_binary
     @testutils.skip_from_python(3)
     def testTypeRoundtripBufferArray(self):
         o1 = buffer("".join(map(chr, range(256))))
@@ -213,7 +209,6 @@ class TypesBasicTests(unittest.TestCase):
         self.assertEqual(type(o1[0]), type(o2[0]))
         self.assertEqual(str(o1[0]), str(o2[0]))
 
-    @skip_if_broken_hex_binary
     @testutils.skip_before_python(3)
     def testTypeRoundtripBytes(self):
         o1 = bytes(range(256))
@@ -225,7 +220,6 @@ class TypesBasicTests(unittest.TestCase):
         o2 = self.execute("select %s;", (o1,))
         self.assertEqual(memoryview, type(o2))
 
-    @skip_if_broken_hex_binary
     @testutils.skip_before_python(3)
     def testTypeRoundtripBytesArray(self):
         o1 = bytes(range(256))
@@ -233,7 +227,6 @@ class TypesBasicTests(unittest.TestCase):
         o2 = self.execute("select %s;", (o1,))
         self.assertEqual(memoryview, type(o2[0]))
 
-    @skip_if_broken_hex_binary
     @testutils.skip_before_python(2, 6)
     def testAdaptBytearray(self):
         o1 = bytearray(range(256))
@@ -258,7 +251,6 @@ class TypesBasicTests(unittest.TestCase):
         else:
             self.assertEqual(memoryview, type(o2))
 
-    @skip_if_broken_hex_binary
     @testutils.skip_before_python(2, 7)
     def testAdaptMemoryview(self):
         o1 = memoryview(bytearray(range(256)))
@@ -335,6 +327,104 @@ class AdaptSubclassTest(unittest.TestCase):
            del psycopg2.extensions.adapters[A, psycopg2.extensions.ISQLQuote]
 
 
+class ByteaParserTest(unittest.TestCase):
+    """Unit test for our bytea format parser."""
+    def setUp(self):
+        try:
+            self._cast = self._import_cast()
+        except Exception, e:
+            self._cast = None
+            self._exc = e
+
+    def _import_cast(self):
+        """Use ctypes to access the C function.
+
+        Raise any sort of error: we just support this where ctypes works as
+        expected.
+        """
+        import ctypes
+        lib = ctypes.cdll.LoadLibrary(psycopg2._psycopg.__file__)
+        cast = lib.typecast_BINARY_cast
+        cast.argtypes = [ctypes.c_char_p, ctypes.c_size_t, ctypes.py_object]
+        cast.restype = ctypes.py_object
+        return cast
+
+    def cast(self, buffer):
+        """Cast a buffer from the output format"""
+        l = buffer and len(buffer) or 0
+        rv = self._cast(buffer, l, None)
+
+        if rv is None:
+            return None
+
+        if sys.version_info[0] < 3:
+            return str(rv)
+        else:
+            return rv.tobytes()
+
+    def test_null(self):
+        rv = self.cast(None)
+        self.assertEqual(rv, None)
+
+    def test_blank(self):
+        rv = self.cast(b(''))
+        self.assertEqual(rv, b(''))
+
+    def test_blank_hex(self):
+        # Reported as problematic in ticket #48
+        rv = self.cast(b('\\x'))
+        self.assertEqual(rv, b(''))
+
+    def test_full_hex(self, upper=False):
+        buf = ''.join(("%02x" % i) for i in range(256))
+        if upper: buf = buf.upper()
+        buf = '\\x' + buf
+        rv = self.cast(b(buf))
+        if sys.version_info[0] < 3:
+            self.assertEqual(rv, ''.join(map(chr, range(256))))
+        else:
+            self.assertEqual(rv, bytes(range(256)))
+
+    def test_full_hex_upper(self):
+        return self.test_full_hex(upper=True)
+
+    def test_full_escaped_octal(self):
+        buf = ''.join(("\\%03o" % i) for i in range(256))
+        rv = self.cast(b(buf))
+        if sys.version_info[0] < 3:
+            self.assertEqual(rv, ''.join(map(chr, range(256))))
+        else:
+            self.assertEqual(rv, bytes(range(256)))
+
+    def test_escaped_mixed(self):
+        import string
+        buf = ''.join(("\\%03o" % i) for i in range(32))
+        buf += string.ascii_letters
+        buf += ''.join('\\' + c for c in string.ascii_letters)
+        buf += '\\\\'
+        rv = self.cast(b(buf))
+        if sys.version_info[0] < 3:
+            tgt = ''.join(map(chr, range(32))) \
+                + string.ascii_letters * 2 + '\\'
+        else:
+            tgt = bytes(range(32)) + \
+                (string.ascii_letters * 2 + '\\').encode('ascii')
+
+        self.assertEqual(rv, tgt)
+
+def skip_if_cant_cast(f):
+    def skip_if_cant_cast_(self, *args, **kwargs):
+        if self._cast is None:
+            return self.skipTest("can't test bytea parser: %s - %s"
+                % (self._exc.__class__.__name__, self._exc))
+
+        return f(self, *args, **kwargs)
+
+    return skip_if_cant_cast_
+
+decorate_all_tests(ByteaParserTest, skip_if_cant_cast)
+
+
 def test_suite():
     return unittest.TestLoader().loadTestsFromName(__name__)