All Downloads are FREE. Search and download functionalities are using the official Maven repository.

com.microsoft.sqlserver.jdbc.TVP Maven / Gradle / Ivy

The newest version!
/*
 * Microsoft JDBC Driver for SQL Server Copyright(c) Microsoft Corporation All rights reserved. This program is made
 * available under the terms of the MIT License. See the LICENSE file in the project root for more information.
 */

package com.microsoft.sqlserver.jdbc;

import java.sql.ResultSet;
import java.sql.ResultSetMetaData;
import java.sql.SQLException;
import java.text.MessageFormat;
import java.util.HashSet;
import java.util.Iterator;
import java.util.LinkedHashMap;
import java.util.Map;
import java.util.Map.Entry;
import java.util.Set;


enum TVPType {
    RESULTSET,
    ISQLSERVERDATARECORD,
    SQLSERVERDATATABLE,
    NULL
}


/**
 * 
 * Implementation of Table-valued parameters which provide an easy way to marshal multiple rows of data from a client
 * application to SQL Server without requiring multiple round trips or special server-side logic for processing the
 * data.
 * 

* You can use table-valued parameters to encapsulate rows of data in a client application and send the data to the * server in a single parameterized command. The incoming data rows are stored in a table variable that can then be * operated on by using Transact-SQL. *

* Column values in table-valued parameters can be accessed using standard Transact-SQL SELECT statements. Table-valued * parameters are strongly typed and their structure is automatically validated. The size of table-valued parameters is * limited only by server memory. *

* You cannot return data in a table-valued parameter. Table-valued parameters are input-only; the OUTPUT keyword is not * supported. */ class TVP { String tvpName; String tvpOwningSchema; String tvpDbName; ResultSet sourceResultSet = null; SQLServerDataTable sourceDataTable = null; Map columnMetadata = null; Iterator> sourceDataTableRowIterator = null; ISQLServerDataRecord sourceRecord = null; TVPType tvpType = null; Set columnNames = null; // MultiPartIdentifierState enum MPIState { MPI_VALUE, MPI_PARSE_NONQUOTE, MPI_LOOK_FOR_SEPARATOR, MPI_LOOK_FOR_NEXT_CHAR_OR_SEPARATOR, MPI_PARSE_QUOTE, MPI_RIGHT_QUOTE, } void initTVP(TVPType type, String tvpPartName) throws SQLServerException { tvpType = type; columnMetadata = new LinkedHashMap<>(); parseTypeName(tvpPartName); } TVP(String tvpPartName) throws SQLServerException { initTVP(TVPType.NULL, tvpPartName); } // Name used in CREATE TYPE TVP(String tvpPartName, SQLServerDataTable tvpDataTable) throws SQLServerException { if (tvpPartName == null) { tvpPartName = tvpDataTable.getTvpName(); } initTVP(TVPType.SQLSERVERDATATABLE, tvpPartName); sourceDataTable = tvpDataTable; sourceDataTableRowIterator = sourceDataTable.getIterator(); populateMetadataFromDataTable(); } TVP(String tvpPartName, ResultSet tvpResultSet) throws SQLServerException { initTVP(TVPType.RESULTSET, tvpPartName); sourceResultSet = tvpResultSet; // Populate TVP metadata from ResultSetMetadata. populateMetadataFromResultSet(); } TVP(String tvpPartName, ISQLServerDataRecord tvpRecord) throws SQLServerException { initTVP(TVPType.ISQLSERVERDATARECORD, tvpPartName); sourceRecord = tvpRecord; columnNames = new HashSet<>(); // Populate TVP metadata from ISQLServerDataRecord. populateMetadataFromDataRecord(); // validate sortOrdinal and throw all relevant exceptions before proceeding validateOrderProperty(); } boolean isNull() { return (TVPType.NULL == tvpType); } Object[] getRowData() throws SQLServerException { if (TVPType.RESULTSET == tvpType) { int colCount = columnMetadata.size(); Object[] rowData = new Object[colCount]; for (int i = 0; i < colCount; i++) { try { /* * for Time types, getting TimeStamp instead of Time, because this value will be converted to String * later on. If the value is a time object, the millisecond would be removed. */ if (java.sql.Types.TIME == sourceResultSet.getMetaData().getColumnType(i + 1)) { rowData[i] = sourceResultSet.getTimestamp(i + 1); } else { rowData[i] = sourceResultSet.getObject(i + 1); } } catch (SQLException e) { throw new SQLServerException(SQLServerException.getErrString("R_unableRetrieveSourceData"), e); } } return rowData; } else if (TVPType.SQLSERVERDATATABLE == tvpType) { Map.Entry rowPair = sourceDataTableRowIterator.next(); return rowPair.getValue(); } else return sourceRecord.getRowData(); } boolean next() throws SQLServerException { if (TVPType.RESULTSET == tvpType) { try { return sourceResultSet.next(); } catch (SQLException e) { throw new SQLServerException(SQLServerException.getErrString("R_unableRetrieveSourceData"), e); } } else if (TVPType.SQLSERVERDATATABLE == tvpType) { return sourceDataTableRowIterator.hasNext(); } else if (null != sourceRecord) { return sourceRecord.next(); } return false; } void populateMetadataFromDataTable() throws SQLServerException { if (null != sourceDataTable) { Map dataTableMetaData = sourceDataTable.getColumnMetadata(); if (null == dataTableMetaData || dataTableMetaData.isEmpty()) { throw new SQLServerException(SQLServerException.getErrString("R_TVPEmptyMetadata"), null); } dataTableMetaData.entrySet() .forEach(e -> columnMetadata.put(e.getKey(), new SQLServerMetaData(e.getValue().columnName, e.getValue().javaSqlType, e.getValue().precision, e.getValue().scale))); } } void populateMetadataFromResultSet() throws SQLServerException { if (null != sourceResultSet) { try { ResultSetMetaData rsmd = sourceResultSet.getMetaData(); for (int i = 0; i < rsmd.getColumnCount(); i++) { SQLServerMetaData columnMetaData = new SQLServerMetaData(rsmd.getColumnName(i + 1), rsmd.getColumnType(i + 1), rsmd.getPrecision(i + 1), rsmd.getScale(i + 1)); columnMetadata.put(i, columnMetaData); } } catch (SQLException e) { throw new SQLServerException(SQLServerException.getErrString("R_unableRetrieveColMeta"), e); } } } void populateMetadataFromDataRecord() throws SQLServerException { if (null != sourceRecord) { if (0 >= sourceRecord.getColumnCount()) { throw new SQLServerException(SQLServerException.getErrString("R_TVPEmptyMetadata"), null); } for (int i = 0; i < sourceRecord.getColumnCount(); i++) { Util.checkDuplicateColumnName(sourceRecord.getColumnMetaData(i + 1).columnName, columnNames); // Make a copy here as we do not want to change user's metadata. SQLServerMetaData metaData = new SQLServerMetaData(sourceRecord.getColumnMetaData(i + 1)); columnMetadata.put(i, metaData); } } } void validateOrderProperty() throws SQLServerException { int columnCount = columnMetadata.size(); boolean[] sortOrdinalSpecified = new boolean[columnCount]; int maxSortOrdinal = -1; int sortCount = 0; for (Entry columnPair : columnMetadata.entrySet()) { SQLServerSortOrder columnSortOrder = columnPair.getValue().sortOrder; int columnSortOrdinal = columnPair.getValue().sortOrdinal; if (SQLServerSortOrder.UNSPECIFIED != columnSortOrder) { // check if there's no way sort order could be monotonically increasing if (columnCount <= columnSortOrdinal) { MessageFormat form = new MessageFormat( SQLServerException.getErrString("R_TVPSortOrdinalGreaterThanFieldCount")); throw new SQLServerException(form.format(new Object[] {columnSortOrdinal, columnPair.getKey()}), null, 0, null); } // Check to make sure we haven't seen this ordinal before if (sortOrdinalSpecified[columnSortOrdinal]) { MessageFormat form = new MessageFormat( SQLServerException.getErrString("R_TVPDuplicateSortOrdinal")); throw new SQLServerException(form.format(new Object[] {columnSortOrdinal}), null, 0, null); } sortOrdinalSpecified[columnSortOrdinal] = true; if (columnSortOrdinal > maxSortOrdinal) maxSortOrdinal = columnSortOrdinal; sortCount++; } } if (0 < sortCount) { // validate monotonically increasing sort order. watch for values outside of the sortCount range. if (maxSortOrdinal >= sortCount) { // there is at least one hole, find the first one int i; for (i = 0; i < sortCount; i++) { if (!sortOrdinalSpecified[i]) break; } MessageFormat form = new MessageFormat(SQLServerException.getErrString("R_TVPMissingSortOrdinal")); throw new SQLServerException(form.format(new Object[] {i}), null, 0, null); } } } void parseTypeName(String name) throws SQLServerException { String leftQuote = "[\""; String rightQuote = "]\""; char separator = '.'; int limit = 3; // DbName, SchemaName, table type name String[] parsedNames = new String[limit]; int stringCount = 0; // index of current string in the buffer if ((null == name) || (0 == name.length())) { MessageFormat form = new MessageFormat(SQLServerException.getErrString("R_invalidTVPName")); Object[] msgArgs = {}; throw new SQLServerException(null, form.format(msgArgs), null, 0, false); } StringBuilder sb = new StringBuilder(name.length()); // String buffer to hold white space used when parsing non-quoted strings 'a b . c d' = 'a b' and 'c d' StringBuilder whitespaceSB = null; // Right quote character to use given the left quote character found. char rightQuoteChar = ' '; MPIState state = MPIState.MPI_VALUE; for (int index = 0; index < name.length(); ++index) { char testchar = name.charAt(index); switch (state) { case MPI_VALUE: int quoteIndex; if (Character.isWhitespace(testchar)) // skip the whitespace continue; else if (testchar == separator) { // If separator was found, but no string was found, initialize the string we are parsing to // Empty. parsedNames[stringCount] = ""; stringCount++; } else if (-1 != (quoteIndex = leftQuote.indexOf(testchar))) { // If we are at left quote, record the corresponding right quote for the left quote rightQuoteChar = rightQuote.charAt(quoteIndex); sb.setLength(0); state = MPIState.MPI_PARSE_QUOTE; } else if (-1 != rightQuote.indexOf(testchar)) { // If we shouldn't see a right quote MessageFormat form = new MessageFormat( SQLServerException.getErrString("R_invalidThreePartName")); throw new SQLServerException(null, form.format(new Object[] {}), null, 0, false); } else { sb.setLength(0); sb.append(testchar); state = MPIState.MPI_PARSE_NONQUOTE; } break; case MPI_PARSE_NONQUOTE: if (testchar == separator) { parsedNames[stringCount] = sb.toString(); // set the currently parsed string stringCount = incrementStringCount(parsedNames, stringCount); state = MPIState.MPI_VALUE; } // Quotes are not valid inside a non-quoted name else if ((-1 != rightQuote.indexOf(testchar)) || (-1 != leftQuote.indexOf(testchar))) { MessageFormat form = new MessageFormat( SQLServerException.getErrString("R_invalidThreePartName")); throw new SQLServerException(null, form.format(new Object[] {}), null, 0, false); } else if (Character.isWhitespace(testchar)) { // If it is Whitespace parsedNames[stringCount] = sb.toString(); // Set the currently parsed string if (null == whitespaceSB) whitespaceSB = new StringBuilder(); whitespaceSB.setLength(0); // start to record the white space, if we are parsing a name like "foo bar" we should return // "foo bar" whitespaceSB.append(testchar); state = MPIState.MPI_LOOK_FOR_NEXT_CHAR_OR_SEPARATOR; } else sb.append(testchar); break; case MPI_LOOK_FOR_NEXT_CHAR_OR_SEPARATOR: if (!Character.isWhitespace(testchar)) { // If it is not whitespace if (testchar == separator) { stringCount = incrementStringCount(parsedNames, stringCount); state = MPIState.MPI_VALUE; } else { // If its not a separator and not whitespace sb.append(whitespaceSB); sb.append(testchar); // Need to set the name here in case the string ends here. parsedNames[stringCount] = sb.toString(); state = MPIState.MPI_PARSE_NONQUOTE; } } else { if (null == whitespaceSB) { whitespaceSB = new StringBuilder(); } whitespaceSB.append(testchar); } break; case MPI_PARSE_QUOTE: // if are on a right quote see if we are escaping the right quote or ending the quoted string if (testchar == rightQuoteChar) state = MPIState.MPI_RIGHT_QUOTE; else sb.append(testchar); // Append what we are currently parsing break; case MPI_RIGHT_QUOTE: if (testchar == rightQuoteChar) { // If the next char is a another right quote then we were escaping the right quote sb.append(testchar); state = MPIState.MPI_PARSE_QUOTE; } else if (testchar == separator) { // If its a separator then record what we've parsed parsedNames[stringCount] = sb.toString(); stringCount = incrementStringCount(parsedNames, stringCount); state = MPIState.MPI_VALUE; } else if (!Character.isWhitespace(testchar)) { // If it is not white space we got problems MessageFormat form = new MessageFormat( SQLServerException.getErrString("R_invalidThreePartName")); throw new SQLServerException(null, form.format(new Object[] {}), null, 0, false); } else { // It is a whitespace character // the following char should be whitespace, separator, or end of string anything else is bad parsedNames[stringCount] = sb.toString(); state = MPIState.MPI_LOOK_FOR_SEPARATOR; } break; case MPI_LOOK_FOR_SEPARATOR: if (!Character.isWhitespace(testchar)) { // If it is not whitespace if (testchar == separator) { // If it is a separator stringCount = incrementStringCount(parsedNames, stringCount); state = MPIState.MPI_VALUE; } else { // not a separator MessageFormat form = new MessageFormat( SQLServerException.getErrString("R_invalidThreePartName")); throw new SQLServerException(null, form.format(new Object[] {}), null, 0, false); } } break; } } if (stringCount > limit - 1) { MessageFormat form = new MessageFormat(SQLServerException.getErrString("R_invalidThreePartName")); throw new SQLServerException(null, form.format(new Object[] {}), null, 0, false); } // Resolve final states after parsing the string switch (state) { case MPI_VALUE: // These states require no extra action case MPI_LOOK_FOR_SEPARATOR: case MPI_LOOK_FOR_NEXT_CHAR_OR_SEPARATOR: break; case MPI_PARSE_NONQUOTE: // Dump what ever was parsed case MPI_RIGHT_QUOTE: parsedNames[stringCount] = sb.toString(); break; case MPI_PARSE_QUOTE: // Invalid Ending States default: MessageFormat form = new MessageFormat(SQLServerException.getErrString("R_invalidThreePartName")); throw new SQLServerException(null, form.format(new Object[] {}), null, 0, false); } if (parsedNames[0] == null) { MessageFormat form = new MessageFormat(SQLServerException.getErrString("R_invalidThreePartName")); throw new SQLServerException(null, form.format(new Object[] {}), null, 0, false); } else { // Shuffle the parsed name, from left justification to right justification, i.e. [a][b][null][null] goes to // [null][null][a][b] int offset = limit - stringCount - 1; if (offset > 0) { for (int x = limit - 1; x >= offset; --x) { parsedNames[x] = parsedNames[x - offset]; parsedNames[x - offset] = null; } } } this.tvpName = parsedNames[2]; this.tvpOwningSchema = parsedNames[1]; this.tvpDbName = parsedNames[0]; } /* * Parsing the multi-part identifier string. parameters: name - string to parse left-quote: set of characters which * are valid quoting characters to initiate a quote right-quote: set of characters which are valid to stop a quote, * array index's correspond to the the left-quote array. separator: separator to use limit: number of names to parse * out removequote:to remove the quotes on the returned string */ private int incrementStringCount(String[] ary, int position) throws SQLServerException { ++position; int limit = ary.length; if (position >= limit) { MessageFormat form = new MessageFormat(SQLServerException.getErrString("R_invalidThreePartName")); throw new SQLServerException(null, form.format(new Object[] {}), null, 0, false); } ary[position] = new String(); return position; } String getTVPName() { return tvpName; } String getDbNameTVP() { return tvpDbName; } String getOwningSchemaNameTVP() { return tvpOwningSchema; } int getTVPColumnCount() { return columnMetadata.size(); } Map getColumnMetadata() { return columnMetadata; } }





© 2015 - 2024 Weber Informatics LLC | Privacy Policy