CKQueryOperation working with big batch - objective-c

I have problem with creating CKQuery operation with big batch of data. My query works with 100 results but after more results query fail, because one thread is bad dispatched or something (libdispatch.dylib`dispatch_group_leave:
) i am lost... any idea?
+ (void) fetchAnswersWithRecordId:(CKRecordID *)recordId completionHandler:(CloudKitCompletionHandler)handler {
NSPredicate *predicate = [NSPredicate predicateWithFormat:#"ANSrecordID == %#", recordId];
CKQuery *query = [[CKQuery alloc] initWithRecordType:ckAnswers predicate:predicate];
CKQueryOperation *operation = [[CKQueryOperation alloc] initWithQuery:query];
CKQueryOperation * __weak weakSelf = operation;
operation.resultsLimit = 300;
NSMutableArray *tmp = [[NSMutableArray alloc] init];
operation.recordFetchedBlock = ^(CKRecord *record) {
[tmp addObject:record];
};
operation.queryCompletionBlock = ^(CKQueryCursor *cursor, NSError *error) {
if (!handler)
return;
NSArray *array = [NSArray arrayWithArray:tmp];
if(cursor != nil) {
CKQueryOperation *newOperation = [[CKQueryOperation alloc] initWithCursor:cursor];
newOperation.recordFetchedBlock = weakSelf.recordFetchedBlock;
newOperation.completionBlock = weakSelf.completionBlock;
[[self publicCloudDatabase] addOperation:newOperation];
} else {
NSLog(#"Results: %lu", [array count]);
dispatch_async(dispatch_get_main_queue(), ^{
handler(array, error);
});
}
};
[[self publicCloudDatabase] addOperation:operation];}

I think your issue lies with the __weak operation and the way you create an operation inside another operation. Here is an example (in swift) of how I do something similar i.e. get additional results, but in a fetch and not a query. Note the use of a instance variable to initialize first time and the use of semi-recursion through GCD dispatch_aync:
private func _fetchRecordChangesFromCloud() {
if !_fetching {
// this is the first and only time this code is called in GCD recusion
// we clean the caches we use to collect the results of the fetch
// so we can then save the record in the correct order so references can be created
_fetchedModifiedRecords = []
_fetchedDeletedRecordIDs = []
// mark fetching has started
_fetching = true
}
let operation = CKFetchRecordChangesOperation(recordZoneID: _customRecordZoneID, previousServerChangeToken: _serverChangeToken)
operation.recordChangedBlock = { (record: CKRecord?) in
if let record = record {
println("Received record to save: \(record)")
self._fetchedModifiedRecords.append(record)
}
}
operation.recordWithIDWasDeletedBlock = { (recordID: CKRecordID?) in
if let recordID = recordID {
println("Received recordID to delete: \(recordID)")
self._fetchedDeletedRecordIDs.append(recordID)
}
}
operation.fetchRecordChangesCompletionBlock = {
(serverChangeToken: CKServerChangeToken?, clientChangeToken: NSData?, error: NSError?) -> Void in
if let error = error {
println("Error in fetching record changes: \(error)")
// try again next sync
self._fetchAfterNextSuccessfullSync = true
self._fetching = false
return
}
// fetched records successfuly
println("fetched records successfuly")
if let serverChangeToken = serverChangeToken {
self._serverChangeToken = serverChangeToken
}
if operation.moreComing {
// we need to create another operation object and do it again
dispatch_async(dispatch_get_global_queue(DISPATCH_QUEUE_PRIORITY_DEFAULT, 0)) {
self._fetchRecordChangesFromCloud()
}
} else {
// we are finally done
// process the fetched records
self._processFetchedRecords()
// save all changes back to persistent store
self._saveBackgroundContext()
// we are done
self._fetching = false
}
}
self._privateDatabase.addOperation(operation)
}

FYI, dispatch_async() is not necessary, this is a memory management issue. The following works properly for a multi-batch fetch:
//
__block CKQueryOperation* enumerateOperationActive = nil;
//
NSPredicate* predicate = [NSPredicate predicateWithValue:TRUE];
CKQuery* query = [[[CKQuery alloc] initWithRecordType:#"myType" predicate:predicate] autorelease];
CKQueryOperation* enumerateOperation = [[[CKQueryOperation alloc] initWithQuery:query] autorelease];
// DEBUG: fetch only 1 record in order to "force" a nested CKQueryOperation cycle
enumerateOperation.resultsLimit = 1;
enumerateOperation.recordFetchedBlock = ^(CKRecord* recordFetched)
{
// ...
};
enumerateOperation.queryCompletionBlock = ^(CKQueryCursor* cursor, NSError* error)
{
if (error)
{
// ...
}
else
{
if (cursor)
{
CKQueryOperation* enumerateOperationNested = [[[CKQueryOperation alloc] initWithCursor:cursor] autorelease];
// DEBUG: fetch only 1 record in order to "force" a doubly-nested CKQueryOperation cycle
enumerateOperationNested.resultsLimit = 1;
enumerateOperationNested.recordFetchedBlock = /* re-used */ enumerateOperationActive.recordFetchedBlock;
enumerateOperationNested.queryCompletionBlock = /* re-used */ enumerateOperationActive.queryCompletionBlock;
// CRITICAL: keep track of the very last (active) operation
enumerateOperationActive = enumerateOperationNested;
[database addOperation:enumerateOperationNested];
}
}
};
//
// CRITICAL: keep track of the very last (active) operation
enumerateOperationActive = enumerateOperation;
[database addOperation:enumerateOperation];
NOTE: if you attempt to access (the original) enumerateOperation.queryCompletionBlock instead of (the very last) enumerateOperationActive.queryCompletionBlock, the operation will never complete.

I liked the recursive solution on the main thread. Below is my solution in Objective C. I'm using a class level var: _recordArray, allocated ahead of time.
- (void) readRecords_Resurs: (CKDatabase *) database query: (CKQuery *) query cursor: (CKQueryCursor *) cursor {
// Send either query or cursor
CKQueryOperation *operation;
if (query != nil) operation = [[CKQueryOperation alloc] initWithQuery: query];
else operation = [[CKQueryOperation alloc] initWithCursor: cursor];
operation.recordFetchedBlock = ^(CKRecord *record) {
[_recordArray addObject:record];
};
operation.queryCompletionBlock = ^(CKQueryCursor *cursor, NSError *error) {
if (cursor == nil || error != nil) {
// Done
dispatch_async(dispatch_get_main_queue(), ^{ [self readRecordsDone: error == nil ? nil : [error localizedDescription]]; });
}
else {
// Get next batch using cursor
dispatch_async(dispatch_get_main_queue(), ^{ [self readRecords_Resurs: database query: nil cursor: cursor]; });
}
};
[database addOperation: operation]; // start
}
- (void) readRecordsDone: (NSString *) errorMsg {
}

My solution is a category that uses two operations but both use the same blocks, and you can supply how many results per request you'd like.
#interface CKDatabase (MH)
/* Convenience method for performing a query receiving the results in batches using multiple network calls. Best use max 400 for cursorResultsLimit otherwise server sometimes exceptions telling you to use max 400. Even using CKQueryOperationMaximumResults can cause this exception. */
- (void)mh_performCursoredQuery:(CKQuery *)query cursorResultsLimit:(int)cursorResultsLimit inZoneWithID:(CKRecordZoneID *)zoneID completionHandler:(void (^)(NSArray /* CKRecord */ *results, NSError *error))completionHandler;
#end
#implementation CKDatabase(MH)
- (void)mh_performCursoredQuery:(CKQuery *)query cursorResultsLimit:(int)cursorResultsLimit inZoneWithID:(CKRecordZoneID *)zoneID completionHandler:(void (^)(NSArray /* CKRecord */ *results, NSError *error))completionHandler{
//holds all the records received.
NSMutableArray* records = [NSMutableArray array];
//this block adds records to the result array
void (^recordFetchedBlock)(CKRecord *record) = ^void(CKRecord *record) {
[records addObject:record];
};
//this is used to allow the block to call itself recurively without causing a retain cycle.
void (^queryCompletionBlock)(CKQueryCursor *cursor, NSError *error)
__block __weak typeof(queryCompletionBlock) weakQueryCompletionBlock;
weakQueryCompletionBlock = queryCompletionBlock = ^void(CKQueryCursor *cursor, NSError *error) {
//if any error at all then end with no results. Note its possible that we got some results,
// and even might have got a cursor. However if there is an error then the cursor doesn't work properly so will just return with no results.
if(error){
completionHandler(nil,error);
}
else if(cursor){
CKQueryOperation* cursorOperation = [[CKQueryOperation alloc] initWithCursor:cursor];
cursorOperation.zoneID = zoneID;
cursorOperation.resultsLimit = cursorResultsLimit;
cursorOperation.recordFetchedBlock = recordFetchedBlock;
cursorOperation.queryCompletionBlock = weakQueryCompletionBlock; // use the weak pointer to prevent retain cycle
//start the recursive operation and return.
[self addOperation:cursorOperation];
}
else{
completionHandler(records,nil);
}
};
//start the initial operation.
CKQueryOperation* queryOperation = [[CKQueryOperation alloc] initWithQuery:query];
queryOperation.zoneID = zoneID;
queryOperation.resultsLimit = cursorResultsLimit;
queryOperation.recordFetchedBlock = recordFetchedBlock;
queryOperation.queryCompletionBlock = queryCompletionBlock;
[self addOperation:queryOperation];
}
#end

Related

How to perform all other tasks once the asynchronous dispatch queue finishes execution

I have a PDFDocument where it has some actions on pages like delete,crop,rotate etc.
So when I click on the delete button and click on save(current thread : main thread)
-(void)save
{
// -- process deleted pages first
for( NSString* deletedPageId in self.deletedPageIdList )
{
[self.pdfCoordinator removePageWithId:deletedPageId];
}
// -- wait for all pages to delete before continuing
[self.pdfCoordinator waitUntilAllOperationsAreFinished];
// few lines of code after this should run only when the above code finishes its execution
// code to save the changes in the pdf to managedObject context
}
The code for removePageWithId:
- (void) removePageWithId:(NSString*)pageId
{
NRMPDFOperation *op = [[NRMPDFOperation alloc] initWithNRMPDF:self
pageId:pageId
selector:#selector(removePageOpImpl:)
args:nil];
[self addOperation:op];
[op release];
}
above code creates an operation and adds it to the operation queue for each deletion of the page
code for removePageOpImpl:
- (NSError *)removePageOpImpl:(NRMPDFOperation *)op
{
NSError* error = [self loadDocument];
if( !error )
{
NSUInteger index = [self pageIndexForId:[op pageId]];
if( index < [self pageCount] )
{
[[self pdfDocument] removePageAtIndex:index];
[[self mutablePageIdList] removeObjectAtIndex:index];
[self updatePageLabelsFromIndex:index];
[self updateChangeCount:NSChangeDone];
self.contentsChanged = YES;
}
else
{
// TODO: error
}
}
return error;
}
In the removePageOpImpl: method the line of code
[[self pdfDocument] removePageAtIndex:index]; internally executing some tasks on main thread(but we are making the main thread to wait until this operation finishes).which causes the deadlock.
I tried to execute the code inside removePageOpImpl: in an asynchronous dispatch queue to avoid the deadlock.below is the code for that
- (NSError *)removePageOpImpl:(NRMPDFOperation *)op
{
NSError* error = [self loadDocument];
if( !error )
{
NSUInteger index = [self pageIndexForId:[op pageId]];
if( index < [self pageCount] )
{
dispatch_async(dispatch_get_main_queue(), ^{
[[self pdfDocument] removePageAtIndex:index];
[[self mutablePageIdList] removeObjectAtIndex:index];
[self updatePageLabelsFromIndex:index];
[self updateChangeCount:NSChangeDone];
self.contentsChanged = YES;
});
}
else
{
// TODO: error
}
}
return error;
}
Now I am out from the deadlock. But another issue is by putting the above code into asynchronous block the code which should run after this tasks is executing before this, because of that my app is not behaving as expected.
Code inside waitUntilAllOperationsAreFinished method
- (void) addOperation:(NRMPDFOperation*)operation
{
[_operationSet addObject:operation];
[[self operationQueue] addOperation:operation];
}
- (void) waitUntilAllOperationsAreFinished
{
[[self operationQueue] waitUntilAllOperationsAreFinished];
}
- (NSOperationQueue *)operationQueue
{
if( !_operationQueue )
{
NSOperationQueue *queue = [[NSOperationQueue alloc] init];
[queue setMaxConcurrentOperationCount:1];
_operationQueue = queue;
}
return _operationQueue;
}
This is how the original Save method looks like :
- (void)saveDocumentWithDelegate:(id)delegate didSaveSelector:(SEL)didSaveSelector contextInfo:(void *)contextInfo
{
// TODO: don't just ignore saveOperation
__block BOOL success = YES;
__block NSError *error = nil;
/* write back field changes */
if ([item hasChangesInEditFieldsetFor:#"values"] )
{
//Some code
}
if( self.isPDFEdited )
{
// -- process deleted pages first
for( NSString* deletedPageId in self.deletedPageIdList )
{
[self.itemPDFCoordinator removePageWithId:deletedPageId];
}
// -- wait for all pages to delete before continuing
[self.itemPDFCoordinator waitUntilAllOperationsAreFinished];
// -- replace the search text any pages we deleted
if( [self.deletedPageIdList count] )
{
[self.item setValue:[self.editPDF string] forKeyPath:#"dict.values._searchText"];
}
NSMutableDictionary* originalRotations = [NSMutableDictionary dictionaryWithCapacity:
[self.itemPDFCoordinator pageCount]];
for( NSString* pageId in self.itemPDFCoordinator.pageIdList )
{
NSInteger rotation = [[self.itemPDFCoordinator pageForId:pageId] rotation];
[originalRotations setObject:[NSNumber numberWithInteger:rotation] forKey:pageId];
}
// -- now process actions on remaining pages (crop, rotate, and convert to b&w)
BOOL didCropAnyPages = NO;
NSMutableArray* convertToBwJobs = [NSMutableArray array];
for( NSString* pageId in [self.pageActionDict allKeys] )
{
NSArray* actions = [self.pageActionDict objectForKey:pageId];
for( NSDictionary* action in actions )
{
NSNumber* rotationNumber = [action objectForKey:#"rotation"];
NSValue* cropRectVal = [action objectForKey:#"cropRect"];
NSNumber* convertToBlackAndWhite = [action objectForKey:#"convertToBlackAndWhite"];
if( rotationNumber )
{
[self.itemPDFCoordinator rotateByDegrees:[rotationNumber integerValue]
forPageWithID:pageId];
}
else if( cropRectVal )
{
[self.itemPDFCoordinator setNormalizedBounds:[cropRectVal rectValue]
forBox:kPDFDisplayBoxCropBox
forPageWithID:pageId];
// -- set a flag so we know to recrop the entire document
didCropAnyPages = YES;
}
else if( [convertToBlackAndWhite boolValue] )
{
NSUInteger pageIndex = [self.itemPDFCoordinator pageIndexForId:pageId];
NRMJob* job = [NRMAppJobFactory convertToBlackAndWhiteJobForItem:self.item
pageIndex:pageIndex];
[convertToBwJobs addObject:job];
}
}
}
// -- reapply crop box to any cropped pages
if( didCropAnyPages )
{
[self.itemPDFCoordinator applyCropBoxToAllPages];
}
[self.itemPDFCoordinator waitUntilAllOperationsAreFinished];
for( NRMJob* job in convertToBwJobs )
{
if( ![[self.masterDocument docjob] addJob:job forItem:self.item error:&error] )
[NSApp presentError:error];
else
[job waitUntilDone];
}
// -- make sure document attributes are updated
NSDictionary *docDict = [self.itemPDFCoordinator documentAttributes];
NSDictionary *newDict = [(NRMItem *)item updateDocumentAttributes:docDict];
if (![newDict isEqualToDictionary:docDict])
[self.itemPDFCoordinator setDocumentAttributes:newDict];
[self.itemPDFCoordinator waitUntilAllOperationsAreFinished];
// -- check if we need to reprocess any pages
for( NSString* pageId in self.itemPDFCoordinator.pageIdList )
{
NSInteger oldRotation = [[originalRotations objectForKey:pageId] integerValue];
NSInteger newRotation = [[self.itemPDFCoordinator pageForId:pageId] rotation];
if( oldRotation != newRotation )
{
// -- if it's an image page and we already have OCR data for it, we should reanalyze
NSUInteger pageIndex = [self.itemPDFCoordinator pageIndexForId:pageId];
BOOL isPageImage = [self.itemPDFCoordinator isImagePageAtIndex:pageIndex DPI:NULL];
if( isPageImage && [item OCRDataForPageIndex:pageIndex] )
{
NRMJob* job = [NRMAppJobFactory reprocessPageJobForItem:self.item
pageIndex:pageIndex];
[[NSNotificationCenter defaultCenter] addObserver:self
selector:#selector(reanalyzeJobFinished:)
name:kNRMJobFinishedNotification
object:job];
success = [[self.masterDocument docjob] addJob:job forItem:self.item error:&error];
if( !success )
{
if( error )
[self presentError:error];
}
//goto bail;
}
}
}
//Force Save of PDF to Disk
[self.itemPDFCoordinator setManagedObjectContext:[item managedObjectContext]];
[self.itemPDFCoordinator saveChanges];
}
if( success )
{
[self updateChangeCount:NSChangeCleared];
if( self.isPDFEdited )
{
// !CLEARSPLIT! please do not remove this comment
[self.item setValue:nil forKeyPath:#"dict.values._splitId"];
if( ![self loadPDFForItem:item error:&error] )
goto bail;
}
}
bail:
if( error )
[self presentError:error];
if( delegate )
{
/* signature:
- (void)document:(NSDocument *)document didSave:(BOOL)didSaveSuccessfully contextInfo:(void *)contextInfo;
*/
objc_msgSend( delegate, didSaveSelector, self, (error ? NO : YES), contextInfo );
}
}
can anyone suggest me how can I get out of this issue.
Rather than calling waitUntilAllOperationsAreFinished (which blocks a thread), I'd suggest specifying an additional completion operation, which you can make dependent upon the other operations finishing:
-(void)save {
NSOperation *completionOperation = [NSBlockOperation blockOperationWithBlock:^{
// whatever you want to do when it's done
}];
for (NSString* deletedPageId in self.deletedPageIdList) {
NSOperation *operation = [self.pdfCoordinator removePageWithId:deletedPageId];
[completionOperation addDependency:operation];
}
// add this completionOperation to a queue, but it won't actually start until the other operations finish
[queue addOperation:completionOperation];
}
Clearly this assumes that you change removePageWithId to return the operation it added to the queue:
- (NSOperation *)removePageWithId:(NSString*)pageId {
NRMPDFOperation *operation = [[NRMPDFOperation alloc] initWithNRMPDF:self
pageId:pageId
selector:#selector(removePageOpImpl:)
args:nil];
[self addOperation:operation];
[operation release];
return operation;
}
This way, you're not blocking any thread waiting for the operations to finish, but simply specify what to do when they do finish.

Parent to child object pass for one to many relationship with multilevel while inserting new child object . Ios Core data

I am having issue with coredata saving in background. I am implementing the following model :
MasterManagedObject (of type NSPrivateQueueConcurrencyType)
MainManagedObjectContext (of type NSMainQueueConcurrencyType & is child of MasterManagedObject)
TemporaryManagedObjectContext (of type NSPrivateQueueConcurrencyType & is child of MainManagedObjectContext)
Code is :
- (NSManagedObjectContext *)masterManagedObjectContext {
if (_masterManagedObjectContext) {
return _masterManagedObjectContext;
}
NSPersistentStoreCoordinator *coordinator = [self storeCoordinator];
if (coordinator != nil) {
dime(#"Here in master context");
_masterManagedObjectContext = [[NSManagedObjectContext alloc] initWithConcurrencyType:NSPrivateQueueConcurrencyType];
[_masterManagedObjectContext setPersistentStoreCoordinator:coordinator];
}
return _masterManagedObjectContext;
}
- (NSManagedObjectContext *)mainManagedObjectContext {
if (_mainManagedObjectContext) {
return _mainManagedObjectContext;
}
_mainManagedObjectContext = [[NSManagedObjectContext alloc] initWithConcurrencyType:NSMainQueueConcurrencyType];
[_mainManagedObjectContext setParentContext:self.masterManagedObjectContext];
return _mainManagedObjectContext;
}
+ (NSManagedObjectContext *)temporaryWorkerContext {
NSManagedObjectContext *tempMOContext = [[NSManagedObjectContext alloc] initWithConcurrencyType:NSPrivateQueueConcurrencyType];
tempMOContext.parentContext = [[DDPersist manager] mainManagedObjectContext];
return tempMOContext;
}
Saving contexts:
+ (void)saveTempContext:(NSManagedObjectContext *)context {
NSError *error;
[context save:&error];
if (!error) {
[[DDPersist manager] saveMainContext];
dime(#"Temp Context Saved");
} else {
dime(#"Temp Context Error = %#",error);
}
}
- (void)saveMainContext {
[[[DDPersist manager] mainManagedObjectContext] performBlock:^{
NSError *error = nil;
[[[DDPersist manager] mainManagedObjectContext] save:&error];
if(!error){
//Write to disk after saving on the main UI context
[[DDPersist manager] saveMasterContext];
dime(#"main Context Saved");
} else {
dime(#"Main Context Error = %#",error);
}
}];
}
- (void)saveMasterContext {
[self.masterManagedObjectContext performBlock:^{
NSError *error = nil;
[self.masterManagedObjectContext save:&error];
if(error){
dime(#"Master Context Saved");
} else {
dime(#"Master Context Error %#", error);
if([NSThread isMainThread]) {
dime(#"Master Context Error NOT ON BACKGROUND CONTEXT! WILL AUTOMATICALLY PERSIST ON MAIN CTX!");
}
}
}];
}
I am using above to create new spaceChecklistItems objects in background thread as below :
//space is parent of spaceCheckListItem with one to many relationship.
__block NSManagedObjectID *spaceObjectID = [space objectID];
//Background thread starts here
[DDPersist performTaskOnBackgroundCtxWithParentChildScheme:^(NSManagedObjectContext *bgCtx) {
Space *localSpace = (Space*)[bgCtx objectWithID:spaceObjectID];
for(NSDictionary * spaceChecklistItemDict in spaceChecklistItems) {
SpaceChecklistItem * spaceChecklistItem = [SpaceChecklistItemService importSpaceChecklistItem:spaceChecklistItemDict space:localSpace];
NSAssert(spaceChecklistItem, #"invalid SpaceChecklistItem at import!");
if(!spaceChecklistItem) continue;
}
[bgCtx obtainPermanentIDsForObjects:bgCtx.insertedObjects.allObjects error:nil];
[DDPersist saveTempContext:bgCtx];
}];
The method(importSpaceChecklistItem) used in the background context is as under :
+ (SpaceChecklistItem*)importSpaceChecklistItem:(NSDictionary*)itemDict space:(Space*)space {
NSNumber *spaceChecklistItemId = [itemDict objectForKey:#"id"];
NSString * inspectionStatus ;
if ([itemDict objectForKey:#"inspectionStatus"]) {
inspectionStatus = [itemDict objectForKey:#"inspectionStatus"];
} else {
inspectionStatus = #"UNDECIDED";
}
NSString * notes = [itemDict objectForKey:#"notes"];
MOC * ctx = space.managedObjectContext;
SpaceChecklistItem * spaceChecklistItem = [NSEntityDescription insertNewObjectForEntityForName:NSStringFromClass([SpaceChecklistItem class])
inManagedObjectContext:ctx];
spaceChecklistItem.spaceChecklistItemId = spaceChecklistItemId;
spaceChecklistItem.space = space;// This is the relationship saving & not working.
spaceChecklistItem.inspectionStatus = inspectionStatus;
spaceChecklistItem.notes=notes;
spaceChecklistItem.sync = #NO;
return spaceChecklistItem;
}
The main issue is performance issue. I want to speedup for loop : for(NSDictionary * spaceChecklistItemDict in spaceChecklistItems) from the above. And want to all the processing into background. This for loop might contain more than 50000 iterations. which usually take time(about 3 minutes) to save into coredata.Data is saving if i use a single thread & keep for loop in the single thread child of main(not the master context) . But this one to many relationship giving me issues & i am struggling with it for a long.
I read many stackoverflow questions & many othe articles . But cant get this sort out. Any help will be appreciated.
Have you run Instruments?
Run the Time Profiler and look at what is taking the most amount of time.
Post that trace to your question so that others can see it as well.

(Objective-c/Mac OSX) How to distinguish managed AD users (AD user create mobile card) from local users on Mac OSX

<\RESOLVED>, Please see the first reply
My mac(10.9) has joined into a AD domain. In my program, I tried to recognize whether the current login user is local account or AD user. I can successfully distinguish them by using the following code.
+ (bool)isLocalUser:(NSString*)user
{
NSError *dirSearchError = nil;
ODRecord *foundUser = findUser(user, &dirSearchError);
if(foundUser !=nil)
{
return YES;
}else
{
return NO;
}
}
ODRecord *findUser(NSString *user, NSError **error)
{
NSLog(#"[MacLogonUI] findUser");
ODNode *searchNode = [ODNode nodeWithSession: [ODSession defaultSession]
type: kODNodeTypeLocalNodes
error: error];
if (searchNode == nil) {
return nil;
}
NSDictionary *nodeInfo = [searchNode nodeDetailsForKeys:nil error:error];
/* query this node for the user record we're interested in.
* We only need one result, which is why maximumResults is set to 1.
*/
ODQuery *userSearch = [ODQuery queryWithNode: searchNode
forRecordTypes: kODRecordTypeUsers
attribute: kODAttributeTypeRecordName
matchType: kODMatchEqualTo
queryValues: user
returnAttributes: kODAttributeTypeStandardOnly
maximumResults: 1
error: error];
if (userSearch == nil) {
return nil;
}
/* For this example we'll use a synchronous search. This could take a while
* so asynchronous searching is preferable.
*/
NSArray *foundRecords = [userSearch resultsAllowingPartial: NO error: error];
if (foundRecords == nil || [foundRecords count] == 0) {
return nil;
}
ODRecord *userRecord = [foundRecords objectAtIndex: 0];
return [[userRecord retain] autorelease];
}
While when the AD user create a mobile card, it is viewed as a managed user(from the System preference -> Users & Groups). The code also recognize this kind of AD user as local. How to deal with this kind of situation?
Do you guys have any idea of this problem?
I have solved this problem by myself. Hope the following code helps:
#import "DasUser.h"
#import <OpenDirectory/OpenDirectory.h>
#import <Collaboration/Collaboration.h>
#implementation DasUser
+ (bool)isLocalUser:(NSString*)user
{
NSError *dirSearchError = nil;
ODRecord *foundUser = findUser(user, &dirSearchError);
if(foundUser !=nil)
{
return YES;
}else
{
return NO;
}
}
ODRecord *findUser(NSString *user, NSError **error)
{
NSLog(#"[MacLogonUI] findUser");
CSIdentityAuthorityRef defaultAuthority = CSGetManagedIdentityAuthority();
CSIdentityClass identityClass = kCSIdentityClassUser;
CSIdentityQueryRef query = CSIdentityQueryCreate(NULL, identityClass, defaultAuthority);
CFErrorRef err = NULL;
CSIdentityQueryExecute(query, 0, &err);
CFArrayRef results = CSIdentityQueryCopyResults(query);
int numResults = CFArrayGetCount(results);
NSMutableArray * managedUsers = [NSMutableArray array];
for (int i = 0; i < numResults; ++i) {
CSIdentityRef identity = (CSIdentityRef)CFArrayGetValueAtIndex(results, i);
CBIdentity * identityObject = [CBIdentity identityWithCSIdentity:identity];
NSString* posixName = [identityObject posixName];
[managedUsers addObject:posixName];
}
CFRelease(results);
CFRelease(query);
ODNode *searchNode = [ODNode nodeWithSession: [ODSession defaultSession]
type: kODNodeTypeLocalNodes
error: error];
if (searchNode == nil) {
return nil;
}
/* query this node for the user record we're interested in.
* We only need one result, which is why maximumResults is set to 1.
*/
ODQuery *userSearch = [ODQuery queryWithNode: searchNode
forRecordTypes: kODRecordTypeUsers
attribute: kODAttributeTypeRecordName
matchType: kODMatchEqualTo
queryValues: user
returnAttributes: kODAttributeTypeStandardOnly
maximumResults: 1
error: error];
if (userSearch == nil) {
return nil;
}
/* For this example we'll use a synchronous search. This could take a while
* so asynchronous searching is preferable.
*/
NSArray *foundRecords = [userSearch resultsAllowingPartial: NO error: error];
if([foundRecords count]>0)
{
NSString *nameStr = [foundRecords[0] recordName];
NSLog(#"[MacLogonUI] findUser nameStr %#", nameStr);
int j;
for( j = 0; j<[managedUsers count]; j++)
{
if([nameStr isEqualToString:managedUsers[j]])
{
break;
}
}
if(j<[managedUsers count])
{
foundRecords = nil;
}
}
if (foundRecords == nil || [foundRecords count] == 0) {
return nil;
}
ODRecord *userRecord = [foundRecords objectAtIndex: 0];
return [[userRecord retain] autorelease];
}
#end
While when network of the mac is disconnected. The managed user can not be listed. Is there anybody has any idea of this?

Using NSProgress with nested NSOperations

I've been investigating NSProgress but have found the existing documentation, class reference and tutorials to be lacking. I'm mainly wondering if my NSProgress is applicable to my use case. The class reference documentation alternatively refers to suboperations or subtasks, I may be mistaken but I interpreted suboperations to mean a case where an NSOperation manages a group of other NSOperations. An example of my use case is as follows:
Create an Upload All Items in Group operation for each group that exists.
Add each of these operations to an NSOperationQueue.
Each Upload All Items in Group operation will create an Upload Item operation for each item in their group. These all get added to an NSOperationQueue managed by the operation.
I would have expected NSProgress to support this, and allow me to propagate progress from the nested operations (Upload Item operation) to the parent operation, and then finally to the main thread and the UI. But I've had difficulty implementing this, it seems as though NSProgress is meant more for long operations that execute all their code on one background thread, but have separate "sections" that make it easy to determine when progress has been made, if this is the case then the use of the term suboperation is a bit misleading as it brings to mind the use of nested NSOperations.
Thank you for any help you can provide, and let me know if additional details are needed.
NSProgress knows nothing about NSOperations -- the two things are orthogonal -- but that doesn't mean it can't be used with them. The idea behind nesting NSProgress "tasks" is that the inner task doesn't know anything about the outer task, and the outer task doesn't need direct access to the inner task's NSProgress to pull in updates for it. I cooked up a little example:
// Outer grouping
NSProgress* DownloadGroupsOfFiles(NSUInteger numGroups, NSUInteger filesPerGroup)
{
// This is the top level NSProgress object
NSProgress* p = [NSProgress progressWithTotalUnitCount: numGroups];
for (NSUInteger i = 0; i < numGroups; ++i)
{
// Whatever DownloadFiles does, it's worth "1 unit" to us.
[p becomeCurrentWithPendingUnitCount: 1];
DownloadFiles(filesPerGroup);
[p resignCurrent];
}
return p;
}
// Inner grouping
void DownloadFiles(NSUInteger numberOfFiles)
{
NSProgress* p = [NSProgress progressWithTotalUnitCount: numberOfFiles];
NSOperationQueue* opQueue = [[NSOperationQueue alloc] init];
// Make the op queue last as long as the NSProgress
objc_setAssociatedObject(p, NULL, opQueue, OBJC_ASSOCIATION_RETAIN);
// For each file...
for (NSUInteger i = 0; i < numberOfFiles; ++i)
{
// Whatever this DownloadOperation does is worth 1 "unit" to us.
[p becomeCurrentWithPendingUnitCount: 1];
// Make the new operation
MyDownloadOperation* op = [[MyDownloadOperation alloc] initWithName: [NSString stringWithFormat: #"File #%#", #(i+1)]];
[opQueue addOperation: op];
[p resignCurrent];
}
}
// And then the DownloadOperation might look like this...
#interface MyDownloadOperation : NSOperation
#property (nonatomic, readonly, copy) NSString* name;
- (id)initWithName: (NSString*)name;
#end
#implementation MyDownloadOperation
{
NSProgress* _progress;
NSString* _name;
}
- (id)initWithName:(NSString *)name
{
if (self = [super init])
{
_name = [name copy];
// Do this in init, so that our NSProgress instance is parented to the current one in the thread that created the operation
_progress = [NSProgress progressWithTotalUnitCount: 1];
}
return self;
}
- (void)dealloc
{
_name = nil;
_progress = nil;
}
- (void)main
{
// Fake like we're doing something that takes some time
// Determine fake size -- call it 768K +- 256K
const NSUInteger size = 512 * 1024 + arc4random_uniform(512*1024);
const NSUInteger avgBytesPerSec = 1024 * 1024;
const NSTimeInterval updatePeriod = 1.0/60.0;
// Make sure all the updates to the NSProgress happen on the main thread
// in case someone is bound to it.
dispatch_async(dispatch_get_main_queue(), ^{
_progress.totalUnitCount = size;
_progress.completedUnitCount = 0;
});
NSUInteger bytesRxd = 0;
do
{
// Sleep for a bit...
usleep(USEC_PER_SEC * updatePeriod);
// "Receive some data"
NSUInteger rxdThisTime = updatePeriod * avgBytesPerSec;
// Never report more than all the bytes
bytesRxd = MIN(bytesRxd + rxdThisTime, size);
// Update on the main thread...
dispatch_async(dispatch_get_main_queue(), ^{
[_progress setCompletedUnitCount: bytesRxd];
});
} while (bytesRxd < size);
}
#end
One thing to note is that if NSProgress is being used to convey status to the UI, then you will want to make sure that every time you update the NSProgress object, you do so from the main thread, otherwise you'll get lots of weird crashes.
Alternately you could just use NSURLConnection to download files, and then have a delegate like this:
#interface MyURLConnectionProgressReporter : NSObject <NSURLConnectionDownloadDelegate>
#property (nonatomic, readwrite, assign) id<NSURLConnectionDownloadDelegate> delegate;
#end
NSProgress* DownloadABunchOfFiles(NSArray* arrayOfURLs)
{
arrayOfURLs = arrayOfURLs.count ? arrayOfURLs : #[ [NSURL URLWithString: #"http://www.google.com"] ];
NSProgress* p = [NSProgress progressWithTotalUnitCount: arrayOfURLs.count];
for (NSURL* url in arrayOfURLs)
{
[p becomeCurrentWithPendingUnitCount: 1];
MyURLConnectionProgressReporter* delegate = [[MyURLConnectionProgressReporter alloc] init];
NSURLConnection* conn = [[NSURLConnection alloc] initWithRequest: [NSURLRequest requestWithURL: url] delegate: delegate];
[conn start];
[p resignCurrent];
}
return p;
}
#implementation MyURLConnectionProgressReporter
{
NSProgress* _progress;
}
static void EnsureMainThread(dispatch_block_t block);
- (id)init
{
if (self = [super init])
{
_progress = [NSProgress progressWithTotalUnitCount: 1];
EnsureMainThread(^{
_progress.kind = NSProgressKindFile;
[_progress setUserInfoObject:NSProgressFileOperationKindDownloading forKey:NSProgressFileOperationKindKey];
});
}
return self;
}
- (id)forwardingTargetForSelector:(SEL)aSelector
{
id retVal = [super forwardingTargetForSelector:aSelector];
if (!retVal && [self.delegate respondsToSelector: _cmd])
{
retVal = self.delegate;
}
return retVal;
}
- (void)p_updateWithTotalBytesWritten:(long long)totalBytesWritten expectedTotalBytes:(long long) expectedTotalBytes
{
// Update our progress on the main thread...
EnsureMainThread(^{
if (!expectedTotalBytes)
_progress.totalUnitCount = -1;
else
_progress.totalUnitCount = MAX(_progress.totalUnitCount, expectedTotalBytes);
_progress.completedUnitCount = totalBytesWritten;
});
}
- (void)connection:(NSURLConnection *)connection didWriteData:(long long)bytesWritten totalBytesWritten:(long long)totalBytesWritten expectedTotalBytes:(long long) expectedTotalBytes
{
// Update our progress
[self p_updateWithTotalBytesWritten: totalBytesWritten expectedTotalBytes: expectedTotalBytes];
// Then call on through to the other delegate
if ([self.delegate respondsToSelector: _cmd])
{
[self.delegate connection:connection didWriteData:bytesWritten totalBytesWritten:totalBytesWritten expectedTotalBytes:expectedTotalBytes];
}
}
- (void)connectionDidResumeDownloading:(NSURLConnection *)connection totalBytesWritten:(long long)totalBytesWritten expectedTotalBytes:(long long) expectedTotalBytes
{
// Update our progress
[self p_updateWithTotalBytesWritten: totalBytesWritten expectedTotalBytes: expectedTotalBytes];
// Then call on through to the other delegate
if ([self.delegate respondsToSelector: _cmd])
{
[self.delegate connectionDidResumeDownloading:connection totalBytesWritten:totalBytesWritten expectedTotalBytes:expectedTotalBytes];
}
}
- (void)connectionDidFinishDownloading:(NSURLConnection *)connection destinationURL:(NSURL *) destinationURL
{
// We're done, so we want (_progress.completedUnitCount == _progress.totalUnitCount)
EnsureMainThread(^{
_progress.completedUnitCount = _progress.totalUnitCount;
});
if ([self.delegate respondsToSelector: _cmd])
{
[self.delegate connectionDidFinishDownloading:connection destinationURL:destinationURL];
}
}
static void EnsureMainThread(dispatch_block_t block)
{
if (!block)
return;
else if ([NSThread isMainThread])
block();
else
dispatch_async(dispatch_get_main_queue(), block);
}
#end
Hope that helps.

Mysterious EXC_BAD_ACCESS with no zombies or uninitialized variables

My app crashes on the following line:
sqlite3_prepare_v2(db, [sql UTF8String], -1, &pStmt, 0);
in the method of the FMDB sqlite wrapper:
- (FMResultSet *)executeQuery:(NSString *)sql withArgumentsInArray:(NSArray*)arrayArgs orVAList:(va_list)args {
if (![self databaseExists]) {
return 0x00;
}
if (inUse) {
[self warnInUse];
return 0x00;
}
[self setInUse:YES];
FMResultSet *rs = nil;
int rc = 0x00;
sqlite3_stmt *pStmt = 0x00;
FMStatement *statement = 0x00;
if (traceExecution && sql) {
NSLog(#"%# executeQuery: %#", self, sql);
}
if (shouldCacheStatements) {
statement = [self cachedStatementForQuery:sql];
pStmt = statement ? [statement statement] : 0x00;
}
int numberOfRetries = 0;
BOOL retry = NO;
if (!pStmt) {
do {
retry = NO;
const char *sqlStatement = [sql UTF8String];
rc = sqlite3_prepare_v2(db, sqlStatement, -1, &pStmt, 0);
if (SQLITE_BUSY == rc || SQLITE_LOCKED == rc) {
retry = YES;
usleep(20);
if (busyRetryTimeout && (numberOfRetries++ > busyRetryTimeout)) {
NSLog(#"%s:%d Database busy (%#)", __FUNCTION__, __LINE__, [self databasePath]);
NSLog(#"Database busy");
sqlite3_finalize(pStmt);
[self setInUse:NO];
return nil;
}
}
else if (SQLITE_OK != rc) {
if (logsErrors) {
NSLog(#"DB Error: %d \"%#\"", [self lastErrorCode], [self lastErrorMessage]);
NSLog(#"DB Query: %#", sql);
#ifndef NS_BLOCK_ASSERTIONS
if (crashOnErrors) {
NSAssert2(false, #"DB Error: %d \"%#\"", [self lastErrorCode], [self lastErrorMessage]);
}
#endif
}
sqlite3_finalize(pStmt);
[self setInUse:NO];
return nil;
}
}
while (retry);
}
id obj;
int idx = 0;
int queryCount = sqlite3_bind_parameter_count(pStmt); // pointed out by Dominic Yu (thanks!)
while (idx < queryCount) {
if (arrayArgs) {
obj = [arrayArgs objectAtIndex:idx];
}
else {
obj = va_arg(args, id);
}
if (traceExecution) {
NSLog(#"obj: %#", obj);
}
idx++;
[self bindObject:obj toColumn:idx inStatement:pStmt];
}
if (idx != queryCount) {
NSLog(#"Error: the bind count is not correct for the # of variables (executeQuery)");
sqlite3_finalize(pStmt);
[self setInUse:NO];
return nil;
}
[statement retain]; // to balance the release below
if (!statement) {
statement = [[FMStatement alloc] init];
[statement setStatement:pStmt];
if (shouldCacheStatements) {
[self setCachedStatement:statement forQuery:sql];
}
}
// the statement gets closed in rs's dealloc or [rs close];
rs = [FMResultSet resultSetWithStatement:statement usingParentDatabase:self];
[rs setQuery:sql];
NSValue *openResultSet = [NSValue valueWithNonretainedObject:rs];
[openResultSets addObject:openResultSet];
statement.useCount = statement.useCount + 1;
[statement release];
[self setInUse:NO];
return rs;
}
The app crashes with EXC_BAD_ACCESS. I have tried to find out why by debugging with NSZombieEnabled and malloc_history, but it does not give me any answers. Also - the debugger tells me that the sql variable has a retain count of a very large number (which is probably because it is a static NSString) - so the EXC_BAD_ACCESS should not be because of the sql object being over-relesed.
Does anyone have any ideas on how to further debug this to find out what the problem is?
Solution: The problem was that my database was accessed by several threads. And even if all threads had synchronized access to the database handle, for sqlite versions prior to 3.3.1 (iOS uses 3.0) you can not safely use the same database handle across threads.
My solution was to create on-demand handles to the database for each thread that tries to access the database, like this:
- (ADatabaseConnection *)databaseConnection {
NSDictionary *dictionary = [[NSThread currentThread] threadDictionary];
NSString *key = #"aDatabaseConnection";
ADatabaseConnection *connection = [dictionary objectForKey:key];
if (connection == nil) {
connection = [[[ADatabaseConnection alloc] initWithDatabase:self] autorelease];
[dictionary setValue:connection forKey:key];
}
return connection;
}
Note that for sqlite versions >= 3.3.1, this is not needed as the same handle can be used across threads.
Another important thing to remember is that even if you use this approach to safely use the same database across threads, it might be wise to synchronize access to the database so that you do not access it simultaneously anyway to avoid database lock errors. I do both, use one handle for each thread and synchronize on the database.
It's not safe to use FMDatabase from multiple threads at the same time- so I've started work on a new class to help you make queries and updates from multiple threads using a pool. Right now it's on a branch, which you can view here:
https://github.com/ccgus/fmdb/tree/threadtests
Read the section titled "Using FMDatabasePool and Thread Safety."