Question

我正在尝试在我的应用中实施搜索。有两个核心数据实体，“Tag”和“DvarTorah”。标签只有一个字符串。 “DvarTorah”具有标题，文本内容和一些其他属性。我正试图找出快速搜索它们的最佳方法。该应用程序附带了大约1200个DvarTorah实体，甚至更多的标签。现在，当我的搜索视图控制器调用viewDidLoad时，我加载了一个NSFetchedResultsController。然后，当用户键入搜索框或更改范围时，我调用一个方法，该方法同时接收范围栏值和搜索词，并过滤我的对象数组。这是看起来的样子：

- (void) filterArrayWithSearchTerm:(NSString *)searchString andScopeIndex:(NSInteger)scopeIndex{

    if ([searchString isEqualToString:@""]) {
        return;
    }    

    NSMutableArray *unfilteredResults = [[[[self.fetchedResultsController sections] objectAtIndex:0] objects] mutableCopy];

    if (self.filteredArray == nil){
        self.filteredArray = [[[NSMutableArray alloc ] init] autorelease];
    }

    [filteredArray removeAllObjects];

    NSPredicate *predicate = [[[NSPredicate alloc] init] autorelease];

    if (scopeIndex == 0) {
        predicate = [NSPredicate predicateWithFormat:@"dvarTorahTitle CONTAINS[cd] %@", searchString];
    }else if (scopeIndex == 1) {
        predicate = [NSPredicate predicateWithFormat:@"searchableContent CONTAINS[cd] %@", [searchString canonicalString]];            
    }else if (scopeIndex == 2){
        predicate = [NSPredicate predicateWithFormat:@"ANY tags.tagText CONTAINS[cd] %@", searchString];
    }else{
        predicate = [NSPredicate predicateWithFormat:@"(ANY tags.tagText CONTAINS[cd] %@) OR (dvarTorahTitle CONTAINS[cd] %@) OR (searchableContent CONTAINS[cd] %@)", searchString,searchString,searchString];
    }

    for (DvarTorah *dvarTorah in unfilteredResults) {
        if ([predicate evaluateWithObject:dvarTorah]) {
            [self.filteredArray addObject:dvarTorah];
        }
    }

    [unfilteredResults release];
}

问题是我的搜索方法非常慢。我知道CONTAINS可能是罪魁祸首，但即使存储了内容的规范版本（作为searchableContent）并尝试进一步优化，搜索也非常缓慢。我怎样才能让它更快？

修改

基于雅各布的初步建议，这是我的新方法：

    if ([searchString isEqualToString:@""]) {
    return;
}

if (self.filteredArray == nil) {
    self.filteredArray = [[[NSMutableArray alloc ] init] autorelease];
}

[filteredArray removeAllObjects];

NSPredicate *predicate = nil;

if (scopeIndex == 0) {
    predicate = [NSPredicate predicateWithFormat:@"dvarTorahTitle CONTAINS[cd] %@", searchString];
}else if (scopeIndex == 1) {
    predicate = [NSPredicate predicateWithFormat:@"searchableContent CONTAINS[cd] %@", [searchString canonicalString]];            
}else if (scopeIndex == 2){
    predicate = [NSPredicate predicateWithFormat:@"ANY tags.tagText CONTAINS[cd] %@", searchString];
}else{
    predicate = [NSPredicate predicateWithFormat:@"(ANY tags.tagText CONTAINS[cd] %@) OR (dvarTorahTitle CONTAINS[cd] %@) OR (searchableContent CONTAINS[cd] %@)", searchString,searchString,searchString];
}

[self.filteredArray addObjectsFromArray:[[[[[self.fetchedResultsController sections] objectAtIndex:0] objects] mutableCopy] filteredArrayUsingPredicate:predicate]];

}

编辑2：

不再复制数组，仍然很慢：

- (void) filterArrayWithSearchTerm:(NSString *)searchString andScopeIndex:(NSInteger)scopeIndex{

    if ([searchString isEqualToString:@""]) {
        return;
    }

    if (self.filteredArray == nil) {
        self.filteredArray = [[[NSMutableArray alloc ] init] autorelease];
    }

    [filteredArray removeAllObjects];

    NSPredicate *predicate = nil;

    if (scopeIndex == 0) {
        predicate = [NSPredicate predicateWithFormat:@"dvarTorahTitle CONTAINS[cd] %@", searchString];
    }else if (scopeIndex == 1) {
        predicate = [NSPredicate predicateWithFormat:@"searchableContent CONTAINS[cd] %@", [searchString canonicalString]];            
    }else if (scopeIndex == 2){
        predicate = [NSPredicate predicateWithFormat:@"ANY tags.tagText CONTAINS[cd] %@", searchString];
    }else{
        predicate = [NSPredicate predicateWithFormat:@"(ANY tags.tagText CONTAINS[cd] %@) OR (dvarTorahTitle CONTAINS[cd] %@) OR (searchableContent CONTAINS[cd] %@)", searchString,searchString,searchString];
    }

    [self.filteredArray addObjectsFromArray:[[[[self.fetchedResultsController sections] objectAtIndex:0] objects] filteredArrayUsingPredicate:predicate]];
}

Answer 1

这里有很多东西可以扼杀CPU周期和内存：

一，你正在对NSFetchedResultsController提取的结果进行可变复制。为什么呢？

其中两个，你在上面的结果上使用for..in构造，并在每个上调用-[NSPredicate evaluateWithObject:]。您可以修改谓词搜索字符串以使用-[NSArray filteredArrayUsingPredicate:]代替，这很可能比您的方法更快。

三，你的predicate变量存在一个相当微妙的问题 - 你总是将它重新分配给开头时自动释放空的其他东西。将其设为默认值nil。

四，正如你所提到的，你的谓词字符串效率很低。我认为你需要做一些叫做索引之类的东西。

有关使用核心数据进行全文搜索的更多信息：

http://developer.apple.com/library/mac/#documentation/cocoa/conceptual/CoreData/Articles/cdPerformance.html

http://cocoawithlove.com/2008/03/testing-core-data-with-very-big.html

http://cocoawithlove.com/2009/11/performance-tests-replacing-core-data.html

http://www.mlsite.net/blog/?page_id=1194

Is SQLite FTS3 still the best way to go for rolling out your full text search?

sqlite Indexing Performance Advice

Full Text Searching in Apple's Core Data Framework

如何优化基于Core Data的搜索？

1 个答案: