案例研究
我有一种方法可以扫描文件系统中的某些条目,一旦找到,就将它们存储为新的对象。
这些对象包括用于条目位置的几个字符串和表示条目状态的标志。
当每个新对象I instanciate占据> 16字节,如果找到1000个条目,我使用(16 * 1000)字节!由于我一次只需要一个实例,我想重复使用一个对象(即一次使用new
)并更新其字段。
我现在如何处理对象:
for (final Object object : scanFor(PATTERN)) {
if (aName.equals(object.getName()) {
... use the object
}
}
我以为我可以使用与ResultSet相同的技术 有什么建议吗?
P.S。 我知道16kb的内存不算什么,但这只是一个例子。
编辑:我看到有人投票结束我的问题,因为它不清楚。我真的不知道还有什么要写的,说实话。 编辑,真实代码:DspPgmRef dspPgmRef = new DspPgmRef("MYLIBRARY", "*ALL", J400Type.ALL);
dspPgmRef.addFilter(DspPgmRefEntryField.REFERENCED_OBJECT, name);
for (final DspPgmRefEntry entry : dspPgmRef.run()) {
if (entry.referencedObject.equals(name)) {
users.add(J400ObjectFactory.get("*LIBL", entry.object));
}
}
DspPgmRef 实施:
public final class DspPgmRef implements J400Command<List<DspPgmRefEntry>>
{
/**
* Rappresenta un valore di ritorno del comando
*/
public static class DspPgmRefEntry
{
public final String library;
public final String object;
public final String referencedLibrary;
public final String referencedObject;
public final String referencedType;
public DspPgmRefEntry(
final String library,
final String object,
final String referencedLibrary,
final String referencedObject,
final String referencedType) {
this.library = library;
this.object = object;
this.referencedLibrary = referencedLibrary;
this.referencedObject = referencedObject;
this.referencedType = referencedType;
}
}
/**
* Campi dei riferimenti ritornati dal comando
*/
public enum DspPgmRefEntryField
{
LIBRARY("WHLIB"),
OBJECT("WHPNAM"),
REFERENCED_LIBRARY("WHLNAM"),
REFERENCED_OBJECT("WHFNAM"),
REFERENCED_TYPE("WHOTYP");
/**
* Nome reale del campo di database
*/
public final String value;
private DspPgmRefEntryField(final String value) {
this.value = value;
}
}
private final String _library;
private final String _object;
private final J400Type _type;
private final Map<DspPgmRefEntryField, String> _filters;
{
_filters = new EnumMap<>(DspPgmRefEntryField.class);
}
public DspPgmRef(final String library, final String object, final J400Type type) {
_library = library;
_object = object;
_type = type;
}
public DspPgmRef(final J400GenericObject object) {
this(object.getLibrary(), object.getName(), object.getType());
}
@Override
public List<DspPgmRefEntry> run() {
final String tempLibrary = J400Util.getRandomAS400String();
final String tempFile = J400Util.getRandomAS400String();
final int filtersSize = _filters.size();
final StringBuilder builder = new StringBuilder(120 + filtersSize * 20);
builder.append("DSPPGMREF PGM(");
builder.append(_library);
builder.append("/");
builder.append(_object);
builder.append(") OUTPUT(*OUTFILE) OBJTYPE(");
builder.append(_type.ASType);
builder.append(") OUTFILE(");
builder.append(tempLibrary);
builder.append("/");
builder.append(tempFile);
builder.append(")");
final J400Connection connection = J400Connection.instance();
connection.createLibrary(tempLibrary, "Libreria temporanea per DSPDBR");
if (!connection.executeCommand(builder.toString()).containsKey("CPF3030")) {
return Collections.EMPTY_LIST;
}
builder.delete(0, builder.length());
builder.append("select distinct WHPNAM, WHLNAM, WHFNAM, WHOTYP from ");
builder.append(tempLibrary);
builder.append(".");
builder.append(tempFile);
if (filtersSize > 0) {
builder.append(" where ");
for (final Entry<DspPgmRefEntryField, String> filter : _filters.entrySet()) {
builder.append(filter.getKey().value);
builder.append(" = '");
builder.append(filter.getValue());
builder.append("' and ");
}
final int stringLenght = builder.length();
builder.delete(stringLenght - 5, stringLenght);
}
PreparedStatement statement = null;
ResultSet result = null;
try {
statement = connection.getConnection().prepareStatement(builder.toString());
result = statement.executeQuery();
final List<DspPgmRefEntry> references = new ArrayList<>(128);
while (result.next()) {
references.add(new DspPgmRefEntry(
_library,
result.getString(DspPgmRefEntryField.OBJECT.value).trim(),
result.getString(DspPgmRefEntryField.REFERENCED_LIBRARY.value).trim(),
result.getString(DspPgmRefEntryField.REFERENCED_OBJECT.value).trim(),
result.getString(DspPgmRefEntryField.REFERENCED_TYPE.value).trim()));
}
return Collections.unmodifiableList(references);
} catch (final SQLException e) {
e.printStackTrace();
} finally {
try {
if (result != null) {
result.close();
}
if (statement != null) {
statement.close();
}
} catch (final SQLException e) {
e.printStackTrace();
}
connection.deleteLibrary(tempLibrary);
}
return Collections.EMPTY_LIST;
}
public void addFilter(final DspPgmRefEntryField field, final String value) {
_filters.put(field, value);
}
public void removeFilter(final DspPgmRefEntryField field) {
_filters.remove(field);
}
public void clearFilters() {
_filters.clear();
}
}
答案 0 :(得分:1)
让我们假设您所建议的是从此代码段中的scanFor
方法返回相同对象实例的性能影响和代码嗅觉:
for (final Object object : scanFor(PATTERN)) {
if (aName.equals(object.getName()) {
... use the object
}
}
根据其余代码的详细信息(例如equals
检查的真实频率和... use the object
中发生的情况),这可能确实可以大大节省分配率。但是,它确实违反了Iterable<T>.iterator().next()
通常理解的行为 - 这是每次都返回一个唯一的对象。这已经在some length before进行了讨论。如果你使用这个技巧测量对你来说很重要的性能提升,它可能仍然值得,但一如既往地衡量。
Java Mission Control包含在(Oracle)JDK中,可以免费用于非商业用途,可以测量实际对象分配率和位置。您可以使用它来验证临时对象是否是整体分配的重要部分(在某些情况下,内联+转义分析可能能够完全消除它们,使您认为的问题无法解决)。