public long getNextSequence(TypeDescriptor typeDescriptor, Session session) { // check if there is an id in the cache ConcurrentLinkedQueue cachedIds = (ConcurrentLinkedQueue) this.typeDescriptorToIdCache.get(typeDescriptor); if (null == cachedIds) { typeDescriptorToIdCache.putIfAbsent(typeDescriptor, new ConcurrentLinkedQueue()); cachedIds = (ConcurrentLinkedQueue) this.typeDescriptorToIdCache.get(typeDescriptor); } Number cachedId = (Number) cachedIds.poll(); if (cachedId == null) { synchronized (cachedIds) { cachedId = (Number) cachedIds.poll(); if (cachedId == null) { List newIds = this.getNextSequenceImpl(typeDescriptor, session); Assert.condition(0 < newIds.size()); // reserve first for own use cachedId = (Number) newIds.remove(0); cachedIds.addAll(newIds); } } } if (trace.isDebugEnabled()) { trace.debug("returning unique ID: " + cachedId.longValue()); } return cachedId.longValue(); }
/* * Making sure that the user input is in range * @param1 Scanner object for input * @param2 lowerbound integer * @param3 upperbound integer * @param4 String that specifies for which attribute/field input is for * printing purposes * @param5 ValType enum that specifies whether the primitive value is * an int or double * @param6 boolean that specifies whether the input is an Enum value * @param7 boolean that specifies whether to return the value if out of * range (false) or not (true) */ public static Number makeSureValInRange( Scanner input, int lowerbound, int upperbound, String inputFor, ValType valType, boolean enumeration, boolean withinRange) { boolean valid = !withinRange; Number val = 0; while (!valid) { System.out.println(inputString(inputFor, null, StringOption.SELECT, enumeration)); val = valType == ValType.INTEGER ? input.nextInt() : input.nextDouble(); if (val.intValue() < lowerbound || val.intValue() > upperbound) { System.out.println(inputString(inputFor, null, StringOption.INCORRECT, enumeration)); } else { valid = true; } } return val; }
/** * 执行统计查询语句,语句的执行结果必须只返回一个数值 * * @param sql * @param params * @return * @throws DBException */ public long stat(String sql, Object... params) throws SQLException { printIn(sql); printIn("params length:" + params.length); for (int i = 0; i < params.length; i++) { printIn("params" + i + ":" + params[i]); } Number num = (Number) queryRunner.query(getConnection(), sql, scalarHandler, params); return (num != null) ? num.longValue() : -1; }
private long getLong(Object value) { if (value instanceof Long) { // Case "value instanceof Number" below is not sufficient for Long: // conversion via double loses precision for values > 2^48. OK for // other types, including int and float. return ((Long) value).longValue(); } else if (value instanceof Number) { Number n = (Number) value; return NumberUtil.round(n.doubleValue()); } else if (value instanceof Boolean) { return (((Boolean) value).booleanValue() ? 1 : 0); } else if (value instanceof String) { try { BigDecimal bd = new BigDecimal(value.toString().trim()); return getLong(bd); } catch (NumberFormatException ex) { throw newInvalidFormat(value); } } else { throw newInvalidType(value); } }
public boolean getBoolean() throws SQLException { int type = JSONTypes.jsonTypes.get(field.getType()); switch (type) { case JSONTypes.JSON_BOOLEAN: return (boolean) jsonObject; case JSONTypes.JSON_NUMBER: Number number = (Number) jsonObject; return !number.equals((Number) 0); case JSONTypes.JSON_STRING: String string = (String) jsonObject; return !string.isEmpty(); case JSONTypes.JSON_MAP: case JSONTypes.JSON_OBJECT: Map map = (Map) jsonObject; return !map.isEmpty(); case JSONTypes.JSON_ARRAY: List list = (List) jsonObject; return !list.isEmpty(); default: return false; } }
public static Object parseType(ResultSet result, Integer i, int type) throws SQLException, IOException, ParseException { logger.trace("i={} type={}", i, type); switch (type) { /** * The JDBC types CHAR, VARCHAR, and LONGVARCHAR are closely related. CHAR represents a * small, fixed-length character string, VARCHAR represents a small, variable-length * character string, and LONGVARCHAR represents a large, variable-length character string. */ case Types.CHAR: case Types.VARCHAR: case Types.LONGVARCHAR: { return result.getString(i); } case Types.NCHAR: case Types.NVARCHAR: case Types.LONGNVARCHAR: { return result.getNString(i); } /** * The JDBC types BINARY, VARBINARY, and LONGVARBINARY are closely related. BINARY * represents a small, fixed-length binary value, VARBINARY represents a small, * variable-length binary value, and LONGVARBINARY represents a large, variable-length * binary value */ case Types.BINARY: case Types.VARBINARY: case Types.LONGVARBINARY: { byte[] b = result.getBytes(i); return b; } /** * The JDBC type ARRAY represents the SQL3 type ARRAY. * * <p>An ARRAY value is mapped to an instance of the Array interface in the Java programming * language. If a driver follows the standard implementation, an Array object logically * points to an ARRAY value on the server rather than containing the elements of the ARRAY * object, which can greatly increase efficiency. The Array interface contains methods for * materializing the elements of the ARRAY object on the client in the form of either an * array or a ResultSet object. */ case Types.ARRAY: { Array arr = result.getArray(i); return arr == null ? null : arr.getArray(); } /** * The JDBC type BIGINT represents a 64-bit signed integer value between * -9223372036854775808 and 9223372036854775807. * * <p>The corresponding SQL type BIGINT is a nonstandard extension to SQL. In practice the * SQL BIGINT type is not yet currently implemented by any of the major databases, and we * recommend that its use be avoided in code that is intended to be portable. * * <p>The recommended Java mapping for the BIGINT type is as a Java long. */ case Types.BIGINT: { Object o = result.getLong(i); return result.wasNull() ? null : o; } /** * The JDBC type BIT represents a single bit value that can be zero or one. * * <p>SQL-92 defines an SQL BIT type. However, unlike the JDBC BIT type, this SQL-92 BIT * type can be used as a parameterized type to define a fixed-length binary string. * Fortunately, SQL-92 also permits the use of the simple non-parameterized BIT type to * represent a single binary digit, and this usage corresponds to the JDBC BIT type. * Unfortunately, the SQL-92 BIT type is only required in "full" SQL-92 and is currently * supported by only a subset of the major databases. Portable code may therefore prefer to * use the JDBC SMALLINT type, which is widely supported. */ case Types.BIT: { try { Object o = result.getInt(i); return result.wasNull() ? null : o; } catch (Exception e) { String exceptionClassName = e.getClass().getName(); // postgresql can not handle boolean, it will throw PSQLException, something like "Bad // value for type int : t" if ("org.postgresql.util.PSQLException".equals(exceptionClassName)) { return "t".equals(result.getString(i)); } throw new IOException(e); } } /** * The JDBC type BOOLEAN, which is new in the JDBC 3.0 API, maps to a boolean in the Java * programming language. It provides a representation of true and false, and therefore is a * better match than the JDBC type BIT, which is either 1 or 0. */ case Types.BOOLEAN: { return result.getBoolean(i); } /** * The JDBC type BLOB represents an SQL3 BLOB (Binary Large Object). * * <p>A JDBC BLOB value is mapped to an instance of the Blob interface in the Java * programming language. If a driver follows the standard implementation, a Blob object * logically points to the BLOB value on the server rather than containing its binary data, * greatly improving efficiency. The Blob interface provides methods for materializing the * BLOB data on the client when that is desired. */ case Types.BLOB: { Blob blob = result.getBlob(i); if (blob != null) { long n = blob.length(); if (n > Integer.MAX_VALUE) { throw new IOException("can't process blob larger than Integer.MAX_VALUE"); } byte[] tab = blob.getBytes(1, (int) n); blob.free(); return tab; } break; } /** * The JDBC type CLOB represents the SQL3 type CLOB (Character Large Object). * * <p>A JDBC CLOB value is mapped to an instance of the Clob interface in the Java * programming language. If a driver follows the standard implementation, a Clob object * logically points to the CLOB value on the server rather than containing its character * data, greatly improving efficiency. Two of the methods on the Clob interface materialize * the data of a CLOB object on the client. */ case Types.CLOB: { Clob clob = result.getClob(i); if (clob != null) { long n = clob.length(); if (n > Integer.MAX_VALUE) { throw new IOException("can't process clob larger than Integer.MAX_VALUE"); } String str = clob.getSubString(1, (int) n); clob.free(); return str; } break; } case Types.NCLOB: { NClob nclob = result.getNClob(i); if (nclob != null) { long n = nclob.length(); if (n > Integer.MAX_VALUE) { throw new IOException("can't process nclob larger than Integer.MAX_VALUE"); } String str = nclob.getSubString(1, (int) n); nclob.free(); return str; } break; } /** * The JDBC type DATALINK, new in the JDBC 3.0 API, is a column value that references a file * that is outside of a data source but is managed by the data source. It maps to the Java * type java.net.URL and provides a way to manage external files. For instance, if the data * source is a DBMS, the concurrency controls it enforces on its own data can be applied to * the external file as well. * * <p>A DATALINK value is retrieved from a ResultSet object with the ResultSet methods * getURL or getObject. If the Java platform does not support the type of URL returned by * getURL or getObject, a DATALINK value can be retrieved as a String object with the method * getString. * * <p>java.net.URL values are stored in a database using the method setURL. If the Java * platform does not support the type of URL being set, the method setString can be used * instead. */ case Types.DATALINK: { return result.getURL(i); } /** * The JDBC DATE type represents a date consisting of day, month, and year. The * corresponding SQL DATE type is defined in SQL-92, but it is implemented by only a subset * of the major databases. Some databases offer alternative SQL types that support similar * semantics. */ case Types.DATE: { try { Date d = result.getDate(i, calendar); return d != null ? formatDate(d.getTime()) : null; } catch (SQLException e) { return null; } } case Types.TIME: { try { Time t = result.getTime(i, calendar); return t != null ? formatDate(t.getTime()) : null; } catch (SQLException e) { return null; } } case Types.TIMESTAMP: { try { Timestamp t = result.getTimestamp(i, calendar); return t != null ? formatDate(t.getTime()) : null; } catch (SQLException e) { // java.sql.SQLException: Cannot convert value '0000-00-00 00:00:00' from column ... to // TIMESTAMP. return null; } } /** * The JDBC types DECIMAL and NUMERIC are very similar. They both represent fixed-precision * decimal values. * * <p>The corresponding SQL types DECIMAL and NUMERIC are defined in SQL-92 and are very * widely implemented. These SQL types take precision and scale parameters. The precision is * the total number of decimal digits supported, and the scale is the number of decimal * digits after the decimal point. For most DBMSs, the scale is less than or equal to the * precision. So for example, the value "12.345" has a precision of 5 and a scale of 3, and * the value ".11" has a precision of 2 and a scale of 2. JDBC requires that all DECIMAL and * NUMERIC types support both a precision and a scale of at least 15. * * <p>The sole distinction between DECIMAL and NUMERIC is that the SQL-92 specification * requires that NUMERIC types be represented with exactly the specified precision, whereas * for DECIMAL types, it allows an implementation to add additional precision beyond that * specified when the type was created. Thus a column created with type NUMERIC(12,4) will * always be represented with exactly 12 digits, whereas a column created with type * DECIMAL(12,4) might be represented by some larger number of digits. * * <p>The recommended Java mapping for the DECIMAL and NUMERIC types is * java.math.BigDecimal. The java.math.BigDecimal type provides math operations to allow * BigDecimal types to be added, subtracted, multiplied, and divided with other BigDecimal * types, with integer types, and with floating point types. * * <p>The method recommended for retrieving DECIMAL and NUMERIC values is * ResultSet.getBigDecimal. JDBC also allows access to these SQL types as simple Strings or * arrays of char. Thus, Java programmers can use getString to receive a DECIMAL or NUMERIC * result. However, this makes the common case where DECIMAL or NUMERIC are used for * currency values rather awkward, since it means that application writers have to perform * math on strings. It is also possible to retrieve these SQL types as any of the Java * numeric types. */ case Types.DECIMAL: case Types.NUMERIC: { BigDecimal bd = null; try { // getBigDecimal() should get obsolete. Most seem to use getString/getObject anyway... bd = result.getBigDecimal(i); } catch (NullPointerException e) { // But is it true? JDBC NPE exists since 13 years? // http://forums.codeguru.com/archive/index.php/t-32443.html // Null values are driving us nuts in JDBC: // http://stackoverflow.com/questions/2777214/when-accessing-resultsets-in-jdbc-is-there-an-elegant-way-to-distinguish-betwee } if (bd == null || result.wasNull()) { return null; } int scale = 2; if (scale >= 0) { bd = bd.setScale(scale, BigDecimal.ROUND_UP); try { long l = bd.longValueExact(); if (Long.toString(l).equals(result.getString(i))) { // convert to long if possible return l; } else { // convert to double (with precision loss) return bd.doubleValue(); } } catch (ArithmeticException e) { return bd.doubleValue(); } } else { return bd.toPlainString(); } } /** * The JDBC type DOUBLE represents a "double precision" floating point number that supports * 15 digits of mantissa. * * <p>The corresponding SQL type is DOUBLE PRECISION, which is defined in SQL-92 and is * widely supported by the major databases. The SQL-92 standard leaves the precision of * DOUBLE PRECISION up to the implementation, but in practice all the major databases * supporting DOUBLE PRECISION support a mantissa precision of at least 15 digits. * * <p>The recommended Java mapping for the DOUBLE type is as a Java double. */ case Types.DOUBLE: { String s = result.getString(i); if (result.wasNull() || s == null) { return null; } NumberFormat format = NumberFormat.getInstance(locale); Number number = format.parse(s); return number.doubleValue(); } /** * The JDBC type FLOAT is basically equivalent to the JDBC type DOUBLE. We provided both * FLOAT and DOUBLE in a possibly misguided attempt at consistency with previous database * APIs. FLOAT represents a "double precision" floating point number that supports 15 digits * of mantissa. * * <p>The corresponding SQL type FLOAT is defined in SQL-92. The SQL-92 standard leaves the * precision of FLOAT up to the implementation, but in practice all the major databases * supporting FLOAT support a mantissa precision of at least 15 digits. * * <p>The recommended Java mapping for the FLOAT type is as a Java double. However, because * of the potential confusion between the double precision SQL FLOAT and the single * precision Java float, we recommend that JDBC programmers should normally use the JDBC * DOUBLE type in preference to FLOAT. */ case Types.FLOAT: { String s = result.getString(i); if (result.wasNull() || s == null) { return null; } NumberFormat format = NumberFormat.getInstance(locale); Number number = format.parse(s); return number.doubleValue(); } /** * The JDBC type JAVA_OBJECT, added in the JDBC 2.0 core API, makes it easier to use objects * in the Java programming language as values in a database. JAVA_OBJECT is simply a type * code for an instance of a class defined in the Java programming language that is stored * as a database object. The type JAVA_OBJECT is used by a database whose type system has * been extended so that it can store Java objects directly. The JAVA_OBJECT value may be * stored as a serialized Java object, or it may be stored in some vendor-specific format. * * <p>The type JAVA_OBJECT is one of the possible values for the column DATA_TYPE in the * ResultSet objects returned by various DatabaseMetaData methods, including getTypeInfo, * getColumns, and getUDTs. The method getUDTs, part of the new JDBC 2.0 core API, will * return information about the Java objects contained in a particular schema when it is * given the appropriate parameters. Having this information available facilitates using a * Java class as a database type. */ case Types.OTHER: case Types.JAVA_OBJECT: { return result.getObject(i); } /** * The JDBC type REAL represents a "single precision" floating point number that supports * seven digits of mantissa. * * <p>The corresponding SQL type REAL is defined in SQL-92 and is widely, though not * universally, supported by the major databases. The SQL-92 standard leaves the precision * of REAL up to the implementation, but in practice all the major databases supporting REAL * support a mantissa precision of at least seven digits. * * <p>The recommended Java mapping for the REAL type is as a Java float. */ case Types.REAL: { String s = result.getString(i); if (result.wasNull() || s == null) { return null; } NumberFormat format = NumberFormat.getInstance(locale); Number number = format.parse(s); return number.doubleValue(); } /** * The JDBC type TINYINT represents an 8-bit integer value between 0 and 255 that may be * signed or unsigned. * * <p>The corresponding SQL type, TINYINT, is currently supported by only a subset of the * major databases. Portable code may therefore prefer to use the JDBC SMALLINT type, which * is widely supported. * * <p>The recommended Java mapping for the JDBC TINYINT type is as either a Java byte or a * Java short. The 8-bit Java byte type represents a signed value from -128 to 127, so it * may not always be appropriate for larger TINYINT values, whereas the 16-bit Java short * will always be able to hold all TINYINT values. */ /** * The JDBC type SMALLINT represents a 16-bit signed integer value between -32768 and 32767. * * <p>The corresponding SQL type, SMALLINT, is defined in SQL-92 and is supported by all the * major databases. The SQL-92 standard leaves the precision of SMALLINT up to the * implementation, but in practice, all the major databases support at least 16 bits. * * <p>The recommended Java mapping for the JDBC SMALLINT type is as a Java short. */ /** * The JDBC type INTEGER represents a 32-bit signed integer value ranging between * -2147483648 and 2147483647. * * <p>The corresponding SQL type, INTEGER, is defined in SQL-92 and is widely supported by * all the major databases. The SQL-92 standard leaves the precision of INTEGER up to the * implementation, but in practice all the major databases support at least 32 bits. * * <p>The recommended Java mapping for the INTEGER type is as a Java int. */ case Types.TINYINT: case Types.SMALLINT: case Types.INTEGER: { try { Integer integer = result.getInt(i); return result.wasNull() ? null : integer; } catch (SQLDataException e) { Long l = result.getLong(i); return result.wasNull() ? null : l; } } case Types.SQLXML: { SQLXML xml = result.getSQLXML(i); return xml != null ? xml.getString() : null; } case Types.NULL: { return null; } /** * The JDBC type DISTINCT field (Types class)>DISTINCT represents the SQL3 type DISTINCT. * * <p>The standard mapping for a DISTINCT type is to the Java type to which the base type of * a DISTINCT object would be mapped. For example, a DISTINCT type based on a CHAR would be * mapped to a String object, and a DISTINCT type based on an SQL INTEGER would be mapped to * an int. * * <p>The DISTINCT type may optionally have a custom mapping to a class in the Java * programming language. A custom mapping consists of a class that implements the interface * SQLData and an entry in a java.util.Map object. */ case Types.DISTINCT: { logger.warn("JDBC type not implemented: {}", type); return null; } /** * The JDBC type STRUCT represents the SQL99 structured type. An SQL structured type, which * is defined by a user with a CREATE TYPE statement, consists of one or more attributes. * These attributes may be any SQL data type, built-in or user-defined. * * <p>The standard mapping for the SQL type STRUCT is to a Struct object in the Java * programming language. A Struct object contains a value for each attribute of the STRUCT * value it represents. * * <p>A STRUCT value may optionally be custom mapped to a class in the Java programming * language, and each attribute in the STRUCT may be mapped to a field in the class. A * custom mapping consists of a class that implements the interface SQLData and an entry in * a java.util.Map object. */ case Types.STRUCT: { logger.warn("JDBC type not implemented: {}", type); return null; } case Types.REF: { logger.warn("JDBC type not implemented: {}", type); return null; } case Types.ROWID: { logger.warn("JDBC type not implemented: {}", type); return null; } default: { logger.warn("unknown JDBC type ignored: {}", type); return null; } } return null; }