[pywin32-checkins] /hgrepo/p/py/pywin32/pywin32: Call PyType_Ready for Cursor_Type ...
OLD project page for the Python extensions for Windows
Brought to you by:
mhammond
Menu
▾
▴
[pywin32-checkins] /hgrepo/p/py/pywin32/pywin32: Call PyType_Ready for Cursor_Type ...
|
From: <pyw...@li...> - 2011-06-24 07:04:00
|
changeset 427c3e914293 in /hgrepo/p/py/pywin32/pywin32 details: http://pywin32.hg.sourceforge.net/hgweb/pywin32/pywin32/hgrepo/p/py/pywin32/pywin32?cmd=changeset;node=427c3e914293 summary: Call PyType_Ready for Cursor_Type and Connection_Type diffstat: win32/src/odbc.cpp | 108 +++++++++++++++++++++++++++--------------------------- 1 files changed, 54 insertions(+), 54 deletions(-) diffs (176 lines): diff -r ade1cd374722 -r 427c3e914293 win32/src/odbc.cpp --- a/win32/src/odbc.cpp Tue Jun 21 22:23:45 2011 -0400 +++ b/win32/src/odbc.cpp Fri Jun 24 02:57:51 2011 -0400 @@ -1168,7 +1168,7 @@ PyObject *sitem = PyObject_Str(item); if (sitem==NULL) rv = 0; - else if PyString_Check(sitem) + else if (PyString_Check(sitem)) rv = ibindString(cur, iCol, sitem); else if PyUnicode_Check(sitem) rv = ibindUnicode(cur, iCol, sitem); @@ -1595,51 +1595,46 @@ return NULL; while (ob) { - SQLLEN cbRequired; - RETCODE rc; - SQLLEN cbRead = 0; - /* Use SQLGetData to retrieve data for blob (or long varchar) type columns. */ - if (ob->bGetData) - { - /* Initialize memory (offsets, etc.) - (use bind_area for buffer, and dynamically allocate in the loop below) */ - cbRequired = ob->vsize; - ob->rcode = 0; - cbRead = 0; /* Count of bytes read (running total and offset into buffer). */ - - /* Loop until SQLGetData tells us that there are no more chunks - of the blob to retrieve. */ - do - { - /* Check to see if bind_area is big enough - if cbRequired > vsize - re-allocate bind_area to cbRequired - set ob->vsize = cbRequired */ - if (cbRequired > ob->vsize) - { - void *pTemp = ob->bind_area; + if (ob->bGetData) + { + /* Use SQLGetData to retrieve data for blob (or long varchar) type columns. + Loop until return code indicates all remaining data fit into buffer. + */ + RETCODE rc; + SQLLEN cbRead = 0; + ob->rcode = 0; + do + { + /* Increase buffer size by cursor chunk size on second and subsequent calls + If not for the SQL Anywhere 5.0 problem (driver version + 5.05.041867), we could probably grow by 50% each time + or the remaining size (as determined by ob->rcode). + Regarding above note, caller can now use cursor.setoutputsize + to work around any such bug in a driver */ + if (ob->rcode){ + void *pTemp = ob->bind_area; + ob->vsize += cur->max_width; /* Some BLOBs can be huge, be paranoid about allowing other threads to run. */ Py_BEGIN_ALLOW_THREADS - ob->bind_area = realloc (ob->bind_area, cbRequired); + ob->bind_area = realloc (ob->bind_area, ob->vsize); Py_END_ALLOW_THREADS if (ob->bind_area == NULL){ PyErr_NoMemory(); + ob->vsize -= cur->max_width; ob->bind_area = pTemp; Py_DECREF(row); return NULL; } - ob->vsize = cbRequired; } - /* rc = GetData( ... , bind_area + offset, vsize - offset, &rcode ) */ Py_BEGIN_ALLOW_THREADS - rc = SQLGetData(cur->hstmt, - ob->pos, - ob->vtype, - (char *)ob->bind_area + cbRead, - ob->vsize - cbRead, - &ob->rcode); + rc = SQLGetData(cur->hstmt, + ob->pos, + ob->vtype, + (char *)ob->bind_area + cbRead, + ob->vsize - cbRead, + &ob->rcode); Py_END_ALLOW_THREADS if (unsuccessful(rc)) { @@ -1647,35 +1642,31 @@ cursorError(cur, _T("SQLGetData")); return NULL; } - /* Returned length is total length remaining including current read. - If length is not known, returns SQL_NO_TOTAL */ - if ((ob->rcode != SQL_NO_TOTAL) && (ob->rcode <= cur->max_width)) + /* Return code can be a negative status code: + SQL_NO_TOTAL if length is not known, SQL_NULL_DATA if nothing to retreive + Otherwise will be total bytes remaining including current read. + */ + if (ob->rcode >= 0 && ob->rcode <= ob->vsize - cbRead) { - /* If we get here, then this should be the last iteration through the loop. */ - cbRead += ob->rcode; + /* If we get here, then this should be the last iteration through the loop. */ + ob->rcode += cbRead; } else { - /* Grow buffer by cursor chunk size for each read. - If not for the SQL Anywhere 5.0 problem (driver version - 5.05.041867), we could probably grow by 50% each time - or the remaining size (as determined by ob->rcode). - Regarding above note, caller can now use cursor.setoutputsize - to work around any such bug in a driver */ - cbRequired += cur->max_width; + cbRead = ob->vsize; /* We want to ignore the intermediate NULL characters SQLGetData() gives us. (silly, silly) */ if (ob->vtype == SQL_C_CHAR) - cbRead = ob->vsize - 1; + cbRead--; else if (ob->vtype == SQL_C_WCHAR) - cbRead = ob->vsize - sizeof(WCHAR); - else - cbRead = ob->vsize; - } + /* Buffer is not guaranteed to be an exact multiple of sizeof(WCHAR), + leaving an extra byte and throwing the next get off by 1. */ + cbRead -= sizeof(WCHAR) + ob->vsize%sizeof(WCHAR); + } - } while (rc == SQL_SUCCESS_WITH_INFO); - } + } while (rc == SQL_SUCCESS_WITH_INFO); + } PyObject *v; if (ob->rcode == SQL_NULL_DATA) @@ -1696,7 +1687,7 @@ { v = ob->copy_fcn( ob->bind_area, - cbRead); + ob->rcode); } if (!v) { @@ -1749,7 +1740,11 @@ } else { - PyList_Append(list, entry); + if (PyList_Append(list, entry) == -1){ + Py_DECREF(list); + Py_DECREF(entry); + return NULL; + } Py_DECREF(entry); } } @@ -1995,6 +1990,11 @@ PYWIN_MODULE_INIT_PREPARE(odbc, globalMethods, "A Python wrapper around the ODBC API."); + if (PyType_Ready(&Cursor_Type) == -1) + PYWIN_MODULE_INIT_RETURN_ERROR; + if (PyType_Ready(&Connection_Type) == -1) + PYWIN_MODULE_INIT_RETURN_ERROR; + // Sql dates are now returned as python's datetime object. // C Api for datetime didn't exist in 2.3, stick to dynamic semantics for now. datetime_module=PyImport_ImportModule("datetime"); |
