Search Results

Search found 2396 results on 96 pages for 'inner'.

Page 67/96 | < Previous Page | 63 64 65 66 67 68 69 70 71 72 73 74  | Next Page >

  • Visual Basic Display Square

    - by user1724157
    Alright I'm currently lost on a particular assignment I have for a class. I've seen many examples of this app but none of them see to help my problem is as follows: Write a Sub procedure "DisplaySquare" to display the solid square. The size should be specified by the integer parameter "size". The character that fills the square should be specified by the string parameter "fillCharacter. Use a For...Next statement nested within another For...Next statement to create the square. The outer For...Next specifies what row is currently being displayed. The inner For...Next appends all the characters that form the row to a display string. So it should come out like as follows: if a user enters "8" and "#" ######## ######## ######## ######## ######## ######## ######## ######## Any help would be appreciated.

    Read the article

  • Charts don't show up when subreport is included in group header in pentaho reporting 3.9.1-GA

    - by user2909808
    There is an issue concerning in sub-reports. I created a bar chart in report header. The sub report(inline) is placed in the Details of main report. In the sub report, the bar chart is placed in the group header. I imported the required parameters to the sub reports from the main report and also i have a sub query for the sub report.The expected output is to show an updated bar chart within each (inner) group of main report. However, the chart is only displayed in the last group occurence, although the chart area itself is allotted in every expected case. Can any one say me, what mistake i did.

    Read the article

  • Is an index required for columns in ON clause?

    - by newbie
    Do I have to create an index on columns referenced in Joins? E.g. SELECT * FROM left_table INNER JOIN right_table ON left_table.foo = right_table.bar WHERE ... Should I create indexes on left_table(foo), right_table(bar), or both? I noticed different results when I used EXPLAIN (Postgresql) with and without indexes and switching around the order of the comparison (right_table.bar = left_table.foo) I know for sure that indexes are used for the left of the WHERE clause but I am wondering whether I need indexes for columns listed in ON clauses.

    Read the article

  • Extendable accessing of sqlite database on android platform

    - by mscriven
    Hi, I am fairly new to the android sdk and databases and have been searching for an answer to this quite some time. I am trying to build an app which has multiple tables within a database. e.g. one for weapons, armours etc. However, my DatabaseManager class which handles all of my table creating, DatabaseHelper inner class and populating of data is creating for an extremely large class requiring high maintenance. Every time I would like to add or remove a table column I need to change quite a few areas of code, - Every reference to the addition of a row in that table with data - The method that the above calls - The method returning all of the database rows - The code in the helper class creating the table - Any specific update methods My question is this: Surely there must be some better way of coding this system, maybe using a database isn't the best way to go, or am i just not used to such large classes having only learned java at university and my largest class consisting of a mere 400-600 lines of code. Thanks for any help!

    Read the article

  • jquery add remove objects

    - by Scarface
    Hey guys quick question, I have an add remove script that will add a unique element and should remove it on a click function but does not. I think it is because the added object is not in the DOM when page is loaded but I am not sure how to fix this. IF anyone has any advice I would greatly appreciate it. $(document).ready(function(){ if (action=='content-change'){ $('#droppable2-inner').empty().append('<div id="content-image"><img id="visual-background2" src=' + src + '></div><div id="drop-content" action="drop-image">x</div>'); } $("#drop-content").click(function() { $('#content-image').remove(); }); })

    Read the article

  • What is better: to delete pointer or set it with a new value?

    - by user63898
    Hi simple question in c++ , say i have a loop and i have function that returns pointer to item so i have to define inner loop pointer so my question is what to do with the pointer inside the loop , delete it ? or to set it with new value is good for example: for(int i =0;i<count();i++) { ptrTmp* ptr = getItemPtr(); // do somthing with the ptr ... // what to do here ? to delete the poinetr or not? delete ptr; // ?? }

    Read the article

  • Spring @Transactional - Can I Override rollbackFor

    - by user475039
    Hi all, I am calling a service which has the following annotation: @Transactional(rollbackFor=ExceptionA.class) public void myMethodA(....) throws ExceptionA { . . } I am calling this method from an other method in another Spring Bean. @Transactional(rollbackFor=ExceptionB.class) public void mainEntryPointMethod(....) throws ExceptionB { . try { myMethodA() } catch (ExceptionA exp) { . } . } My problem is that if myMethodA throws an exception, my transaction (which is passed from mainEntryPointMethod - myMethodA by default propagation) will be marked for rollback. Is there a way in which the 'rollbackFor' for the inner method can be overriden? Thanks in advance Chris

    Read the article

  • Clarifying... So Background Jobs don't Tie Up Application Resources (in Rails)?

    - by viatropos
    I'm trying to get a better grasp of the inner workings of background jobs and how they improve performance. I understand that the goal is to have the application return a response to the user as fast as it can, so you don't want to, say, parse a huge feed that would take 10 seconds because it would prevent the application from being able to process any other requests. So it's recommended to put any operations that take more than say 500ms to execute, into a queued background job. What I don't understand is, doesn't that just delay the same problem? I know the user who invoked that background job will get an immediate response, but what if another user comes right when that background job starts (and it takes 10 seconds to finish), wont that user have to wait? Or is the main issue that, requests are the only thing that can happen one-at-a-time, while on the other hand a request can start while one+ background jobs are in the middle of running? Is that correct?

    Read the article

  • MySQL join not returning rows

    - by John
    I'm attempting to create an anti-bruteforcer for the login page on a website. Unfortunately, my query is not working as expected. I would like to test how many times an IP address has attempted to login, and also return the ID of the user for my next step in the login process. However, I'm having a problem with the query... for one thing, this would only return rows if it was the same user as they had been trying to login to before. I need it to be any user. Secondly, regardless of whether I use LEFT JOIN, RIGHT JOIN, INNER JOIN or JOIN, it will not return the user's ID unless there is a row for the user in login_attempts. SELECT COUNT(`la`.`id`), `u`.`id` FROM `users` AS `u` LEFT JOIN `login_attempts` AS `la` ON `u`.`id` = `la`.`user_id` WHERE `u`.`username` = 'admin' AND `la`.`ip_address` = '127.0.0.1' AND `la`.`timestamp` >= '1'

    Read the article

  • Multiple &(AND) fails in query

    - by N e w B e e
    here is my query $sql = 'SELECT * FROM Orders INNER JOIN [Order Details] ON Orders.OrderNumber = [Order Details].OrderNumber WHERE Orders.CartID =2 AND [Order Details].Option10 Is Null AND [Order Details].Status="Shipped"'; this queries when entered in MS_Access sql view, returns the correct results, but when I copy and paste the same query in my php script, it fails and gives the error Too few parameters, expected 1... although data is there, query is working in access... Please note if I omitted on AND condition, it works eg if I removed shipped conidtion or is null condition, it works then too.. any hint? whats wrong with it?? any help?thanks

    Read the article

  • Sort vector<int>(n) in O(n) time using O(m) space?

    - by Adam
    I have a vector<unsigned int> vec of size n. Each element in vec is in the range [0, m], no duplicates, and I want to sort vec. Is it possible to do better than O(n log n) time if you're allowed to use O(m) space? In the average case m is much larger than n, in the worst case m == n. Ideally I want something O(n). I get the feeling that there's a bucket sort-ish way to do this: unsigned int aux[m]; aux[vec[i]] = i; Somehow extract the permutation and permute vec. I'm stuck on how to do 3. In my application m is on the order of 16k. However this sort is in the inner loops and accounts for a significant portion of my runtime.

    Read the article

  • You have an error in your SQL syntax; check the manual that corresponds to your MySQL

    - by LuisEValencia
    I am trying to run a mysql query to find all occurences of a text. I have a syntax error but dont know where or how to fix it I am using sqlyog to execute this script DECLARE @url VARCHAR(255) SET @url = '1720' SELECT 'select * from ' + RTRIM(tbl.name) + ' where ' + RTRIM(col.name) + ' like %' + RTRIM(@url) + '%' FROM sysobjects tbl INNER JOIN syscolumns col ON tbl.id = col.id AND col.xtype IN (167, 175, 231, 239) -- (n)char and (n)varchar, there may be others to include AND col.length > 30 -- arbitrary min length into which you might store a URL WHERE tbl.type = 'U' -- user defined table 1 queries executed, 0 success, 1 errors, 0 warnings Query: declare @url varchar(255) set @url = '1720' select 'select * from ' + rtrim(tbl.name) + ' where ' + rtrim(col.name) + ' like %' ... Error Code: 1064 You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version for the right syntax to use near 'declare @url varchar(255)

    Read the article

  • IE 8 and jquery problem

    - by Hellnar
    I have such div: <div class="box"> <div id="myid" style="display:hidden;"> <p>some stuff here</p></div> </div> when I do $("myid").slideToggle("slow"); , if the content of has bigger height over box, my content overflows from the bottom of the "box" class, weirdly this doesn't happen with other browsers but IE8 How can I fix this issue so that when "myid" starts to display content, height of the .box resizes itself to fit the inner div.

    Read the article

  • in c++ what is bettr to delete poiner or set it with new value?

    - by user63898
    Hi simple question in c++ , say i have a loop and i have function that returns pointer to item so i have to define inner loop pointer so my question is what to do with the pointer inside the loop , delete it ? or to set it with new value is good for example: for(int i =0;i<count();i++) { ptrTmp* ptr = getItemPtr(); // do somthing with the ptr ... // what to do here ? to delete the poinetr or not? delete ptr; // ?? }

    Read the article

  • mysql query for change in values in a logging table

    - by kiasectomondo
    I have a table like this: Index , PersonID , ItemCount , UnixTimeStamp 1 , 1 , 1 , 1296000000 2 , 1 , 2 , 1296000100 3 , 2 , 4 , 1296003230 4 , 2 , 6 , 1296093949 5 , 1 , 0 , 1296093295 Time and index always go up. Its basically a logging table to log the itemcount each time it changes. I get the most recent ItemCount for each Person like this: SELECT * FROM table a INNER JOIN ( SELECT MAX(index) as i FROM table GROUP BY PersonID) b ON a.index = b.i; What I want to do is get get the most recent record for each PersonID that is at least 24 hours older than the most recent record for each Person ID. Then I want to take the difference in ItemCount between these two to get a change in itemcount for each person over the last 24 hours: personID ChangeInItemCountOverAtLeast24Hours 1 3 2 -11 3 6 Im sort of stuck with what to do next. How can I join another itemcount based on latest adjusted timestamp of individual rows?

    Read the article

  • Selecting only the entries that have a distinct combination of values?

    - by Theodore E O'Neal
    I have a table, links1, that has the columns headers CardID and AbilityID, that looks like this: CardID | AbilityID 1001 | 1 1001 | 2 1001 | 3 1002 | 2 1002 | 3 1002 | 4 1003 | 3 1003 | 4 1003 | 5 What I want is to be able to return all the CardID that that have two specific AbilityID. For example: If I choose 1 and 2, it returns 1001. If I choose 3 and 4, it returns 1002 and 1003. Is it possible to do this with only one table, or will I need to create an identical table and do an INNER JOIN on those?

    Read the article

  • SQL SERVER – Difference Between DATETIME and DATETIME2

    - by pinaldave
    Yesterday I have written a very quick blog post on SQL SERVER – Difference Between GETDATE and SYSDATETIME and I got tremendous response for the same. I suggest you read that blog post before continuing this blog post today. I had asked people to honestly take part and share their view about above two system function. There are few emails as well few comments on the blog post asking question how did I come to know the difference between the same. The answer is real world issues. I was called in for performance tuning consultancy where I was asked very strange question by one developer. Here is the situation he was facing. System had a single table with two different column of datetime. One column was datelastmodified and second column was datefirstmodified. One of the column was DATETIME and another was DATETIME2. Developer was populating them with SYSDATETIME respectively. He was always thinking that the value inserted in the table will be the same. This table was only accessed by INSERT statement and there was no updates done over it in application.One fine day he ran distinct on both of this column and was in for surprise. He always thought that both of the table will have same data, but in fact they had very different data. He presented this scenario to me. I said this can not be possible but when looked at the resultset, I had to agree with him. Here is the simple script generated to demonstrate the problem he was facing. This is just a sample of original table. DECLARE @Intveral INT SET @Intveral = 10000 CREATE TABLE #TimeTable (FirstDate DATETIME, LastDate DATETIME2) WHILE (@Intveral > 0) BEGIN INSERT #TimeTable (FirstDate, LastDate) VALUES (SYSDATETIME(), SYSDATETIME()) SET @Intveral = @Intveral - 1 END GO SELECT COUNT(DISTINCT FirstDate) D_GETDATE, COUNT(DISTINCT LastDate) D_SYSGETDATE FROM #TimeTable GO SELECT DISTINCT a.FirstDate, b.LastDate FROM #TimeTable a INNER JOIN #TimeTable b ON a.FirstDate = b.LastDate GO SELECT * FROM #TimeTable GO DROP TABLE #TimeTable GO Let us see the resultset. You can clearly see from result that SYSDATETIME() does not populate the same value in the both of the field. In fact the value is either rounded down or rounded up in the field which is DATETIME. Event though we are populating the same value, the values are totally different in both the column resulting the SELF JOIN fail and display different DISTINCT values. The best policy is if you are using DATETIME use GETDATE() and if you are suing DATETIME2 use SYSDATETIME() to populate them with current date and time to accurately address the precision. As DATETIME2 is introduced in SQL Server 2008, above script will only work with SQL SErver 2008 and later versions. I hope I have answered few questions asked yesterday. Reference: Pinal Dave (http://www.SQLAuthority.com) Filed under: Pinal Dave, SQL, SQL Authority, SQL DateTime, SQL Optimization, SQL Performance, SQL Query, SQL Scripts, SQL Server, SQL Tips and Tricks, T SQL, Technology

    Read the article

  • Improving Partitioned Table Join Performance

    - by Paul White
    The query optimizer does not always choose an optimal strategy when joining partitioned tables. This post looks at an example, showing how a manual rewrite of the query can almost double performance, while reducing the memory grant to almost nothing. Test Data The two tables in this example use a common partitioning partition scheme. The partition function uses 41 equal-size partitions: CREATE PARTITION FUNCTION PFT (integer) AS RANGE RIGHT FOR VALUES ( 125000, 250000, 375000, 500000, 625000, 750000, 875000, 1000000, 1125000, 1250000, 1375000, 1500000, 1625000, 1750000, 1875000, 2000000, 2125000, 2250000, 2375000, 2500000, 2625000, 2750000, 2875000, 3000000, 3125000, 3250000, 3375000, 3500000, 3625000, 3750000, 3875000, 4000000, 4125000, 4250000, 4375000, 4500000, 4625000, 4750000, 4875000, 5000000 ); GO CREATE PARTITION SCHEME PST AS PARTITION PFT ALL TO ([PRIMARY]); There two tables are: CREATE TABLE dbo.T1 ( TID integer NOT NULL IDENTITY(0,1), Column1 integer NOT NULL, Padding binary(100) NOT NULL DEFAULT 0x,   CONSTRAINT PK_T1 PRIMARY KEY CLUSTERED (TID) ON PST (TID) );   CREATE TABLE dbo.T2 ( TID integer NOT NULL, Column1 integer NOT NULL, Padding binary(100) NOT NULL DEFAULT 0x,   CONSTRAINT PK_T2 PRIMARY KEY CLUSTERED (TID, Column1) ON PST (TID) ); The next script loads 5 million rows into T1 with a pseudo-random value between 1 and 5 for Column1. The table is partitioned on the IDENTITY column TID: INSERT dbo.T1 WITH (TABLOCKX) (Column1) SELECT (ABS(CHECKSUM(NEWID())) % 5) + 1 FROM dbo.Numbers AS N WHERE n BETWEEN 1 AND 5000000; In case you don’t already have an auxiliary table of numbers lying around, here’s a script to create one with 10 million rows: CREATE TABLE dbo.Numbers (n bigint PRIMARY KEY);   WITH L0 AS(SELECT 1 AS c UNION ALL SELECT 1), L1 AS(SELECT 1 AS c FROM L0 AS A CROSS JOIN L0 AS B), L2 AS(SELECT 1 AS c FROM L1 AS A CROSS JOIN L1 AS B), L3 AS(SELECT 1 AS c FROM L2 AS A CROSS JOIN L2 AS B), L4 AS(SELECT 1 AS c FROM L3 AS A CROSS JOIN L3 AS B), L5 AS(SELECT 1 AS c FROM L4 AS A CROSS JOIN L4 AS B), Nums AS(SELECT ROW_NUMBER() OVER (ORDER BY (SELECT NULL)) AS n FROM L5) INSERT dbo.Numbers WITH (TABLOCKX) SELECT TOP (10000000) n FROM Nums ORDER BY n OPTION (MAXDOP 1); Table T1 contains data like this: Next we load data into table T2. The relationship between the two tables is that table 2 contains ‘n’ rows for each row in table 1, where ‘n’ is determined by the value in Column1 of table T1. There is nothing particularly special about the data or distribution, by the way. INSERT dbo.T2 WITH (TABLOCKX) (TID, Column1) SELECT T.TID, N.n FROM dbo.T1 AS T JOIN dbo.Numbers AS N ON N.n >= 1 AND N.n <= T.Column1; Table T2 ends up containing about 15 million rows: The primary key for table T2 is a combination of TID and Column1. The data is partitioned according to the value in column TID alone. Partition Distribution The following query shows the number of rows in each partition of table T1: SELECT PartitionID = CA1.P, NumRows = COUNT_BIG(*) FROM dbo.T1 AS T CROSS APPLY (VALUES ($PARTITION.PFT(TID))) AS CA1 (P) GROUP BY CA1.P ORDER BY CA1.P; There are 40 partitions containing 125,000 rows (40 * 125k = 5m rows). The rightmost partition remains empty. The next query shows the distribution for table 2: SELECT PartitionID = CA1.P, NumRows = COUNT_BIG(*) FROM dbo.T2 AS T CROSS APPLY (VALUES ($PARTITION.PFT(TID))) AS CA1 (P) GROUP BY CA1.P ORDER BY CA1.P; There are roughly 375,000 rows in each partition (the rightmost partition is also empty): Ok, that’s the test data done. Test Query and Execution Plan The task is to count the rows resulting from joining tables 1 and 2 on the TID column: SET STATISTICS IO ON; DECLARE @s datetime2 = SYSUTCDATETIME();   SELECT COUNT_BIG(*) FROM dbo.T1 AS T1 JOIN dbo.T2 AS T2 ON T2.TID = T1.TID;   SELECT DATEDIFF(Millisecond, @s, SYSUTCDATETIME()); SET STATISTICS IO OFF; The optimizer chooses a plan using parallel hash join, and partial aggregation: The Plan Explorer plan tree view shows accurate cardinality estimates and an even distribution of rows across threads (click to enlarge the image): With a warm data cache, the STATISTICS IO output shows that no physical I/O was needed, and all 41 partitions were touched: Running the query without actual execution plan or STATISTICS IO information for maximum performance, the query returns in around 2600ms. Execution Plan Analysis The first step toward improving on the execution plan produced by the query optimizer is to understand how it works, at least in outline. The two parallel Clustered Index Scans use multiple threads to read rows from tables T1 and T2. Parallel scan uses a demand-based scheme where threads are given page(s) to scan from the table as needed. This arrangement has certain important advantages, but does result in an unpredictable distribution of rows amongst threads. The point is that multiple threads cooperate to scan the whole table, but it is impossible to predict which rows end up on which threads. For correct results from the parallel hash join, the execution plan has to ensure that rows from T1 and T2 that might join are processed on the same thread. For example, if a row from T1 with join key value ‘1234’ is placed in thread 5’s hash table, the execution plan must guarantee that any rows from T2 that also have join key value ‘1234’ probe thread 5’s hash table for matches. The way this guarantee is enforced in this parallel hash join plan is by repartitioning rows to threads after each parallel scan. The two repartitioning exchanges route rows to threads using a hash function over the hash join keys. The two repartitioning exchanges use the same hash function so rows from T1 and T2 with the same join key must end up on the same hash join thread. Expensive Exchanges This business of repartitioning rows between threads can be very expensive, especially if a large number of rows is involved. The execution plan selected by the optimizer moves 5 million rows through one repartitioning exchange and around 15 million across the other. As a first step toward removing these exchanges, consider the execution plan selected by the optimizer if we join just one partition from each table, disallowing parallelism: SELECT COUNT_BIG(*) FROM dbo.T1 AS T1 JOIN dbo.T2 AS T2 ON T2.TID = T1.TID WHERE $PARTITION.PFT(T1.TID) = 1 AND $PARTITION.PFT(T2.TID) = 1 OPTION (MAXDOP 1); The optimizer has chosen a (one-to-many) merge join instead of a hash join. The single-partition query completes in around 100ms. If everything scaled linearly, we would expect that extending this strategy to all 40 populated partitions would result in an execution time around 4000ms. Using parallelism could reduce that further, perhaps to be competitive with the parallel hash join chosen by the optimizer. This raises a question. If the most efficient way to join one partition from each of the tables is to use a merge join, why does the optimizer not choose a merge join for the full query? Forcing a Merge Join Let’s force the optimizer to use a merge join on the test query using a hint: SELECT COUNT_BIG(*) FROM dbo.T1 AS T1 JOIN dbo.T2 AS T2 ON T2.TID = T1.TID OPTION (MERGE JOIN); This is the execution plan selected by the optimizer: This plan results in the same number of logical reads reported previously, but instead of 2600ms the query takes 5000ms. The natural explanation for this drop in performance is that the merge join plan is only using a single thread, whereas the parallel hash join plan could use multiple threads. Parallel Merge Join We can get a parallel merge join plan using the same query hint as before, and adding trace flag 8649: SELECT COUNT_BIG(*) FROM dbo.T1 AS T1 JOIN dbo.T2 AS T2 ON T2.TID = T1.TID OPTION (MERGE JOIN, QUERYTRACEON 8649); The execution plan is: This looks promising. It uses a similar strategy to distribute work across threads as seen for the parallel hash join. In practice though, performance is disappointing. On a typical run, the parallel merge plan runs for around 8400ms; slower than the single-threaded merge join plan (5000ms) and much worse than the 2600ms for the parallel hash join. We seem to be going backwards! The logical reads for the parallel merge are still exactly the same as before, with no physical IOs. The cardinality estimates and thread distribution are also still very good (click to enlarge): A big clue to the reason for the poor performance is shown in the wait statistics (captured by Plan Explorer Pro): CXPACKET waits require careful interpretation, and are most often benign, but in this case excessive waiting occurs at the repartitioning exchanges. Unlike the parallel hash join, the repartitioning exchanges in this plan are order-preserving ‘merging’ exchanges (because merge join requires ordered inputs): Parallelism works best when threads can just grab any available unit of work and get on with processing it. Preserving order introduces inter-thread dependencies that can easily lead to significant waits occurring. In extreme cases, these dependencies can result in an intra-query deadlock, though the details of that will have to wait for another time to explore in detail. The potential for waits and deadlocks leads the query optimizer to cost parallel merge join relatively highly, especially as the degree of parallelism (DOP) increases. This high costing resulted in the optimizer choosing a serial merge join rather than parallel in this case. The test results certainly confirm its reasoning. Collocated Joins In SQL Server 2008 and later, the optimizer has another available strategy when joining tables that share a common partition scheme. This strategy is a collocated join, also known as as a per-partition join. It can be applied in both serial and parallel execution plans, though it is limited to 2-way joins in the current optimizer. Whether the optimizer chooses a collocated join or not depends on cost estimation. The primary benefits of a collocated join are that it eliminates an exchange and requires less memory, as we will see next. Costing and Plan Selection The query optimizer did consider a collocated join for our original query, but it was rejected on cost grounds. The parallel hash join with repartitioning exchanges appeared to be a cheaper option. There is no query hint to force a collocated join, so we have to mess with the costing framework to produce one for our test query. Pretending that IOs cost 50 times more than usual is enough to convince the optimizer to use collocated join with our test query: -- Pretend IOs are 50x cost temporarily DBCC SETIOWEIGHT(50);   -- Co-located hash join SELECT COUNT_BIG(*) FROM dbo.T1 AS T1 JOIN dbo.T2 AS T2 ON T2.TID = T1.TID OPTION (RECOMPILE);   -- Reset IO costing DBCC SETIOWEIGHT(1); Collocated Join Plan The estimated execution plan for the collocated join is: The Constant Scan contains one row for each partition of the shared partitioning scheme, from 1 to 41. The hash repartitioning exchanges seen previously are replaced by a single Distribute Streams exchange using Demand partitioning. Demand partitioning means that the next partition id is given to the next parallel thread that asks for one. My test machine has eight logical processors, and all are available for SQL Server to use. As a result, there are eight threads in the single parallel branch in this plan, each processing one partition from each table at a time. Once a thread finishes processing a partition, it grabs a new partition number from the Distribute Streams exchange…and so on until all partitions have been processed. It is important to understand that the parallel scans in this plan are different from the parallel hash join plan. Although the scans have the same parallelism icon, tables T1 and T2 are not being co-operatively scanned by multiple threads in the same way. Each thread reads a single partition of T1 and performs a hash match join with the same partition from table T2. The properties of the two Clustered Index Scans show a Seek Predicate (unusual for a scan!) limiting the rows to a single partition: The crucial point is that the join between T1 and T2 is on TID, and TID is the partitioning column for both tables. A thread that processes partition ‘n’ is guaranteed to see all rows that can possibly join on TID for that partition. In addition, no other thread will see rows from that partition, so this removes the need for repartitioning exchanges. CPU and Memory Efficiency Improvements The collocated join has removed two expensive repartitioning exchanges and added a single exchange processing 41 rows (one for each partition id). Remember, the parallel hash join plan exchanges had to process 5 million and 15 million rows. The amount of processor time spent on exchanges will be much lower in the collocated join plan. In addition, the collocated join plan has a maximum of 8 threads processing single partitions at any one time. The 41 partitions will all be processed eventually, but a new partition is not started until a thread asks for it. Threads can reuse hash table memory for the new partition. The parallel hash join plan also had 8 hash tables, but with all 5,000,000 build rows loaded at the same time. The collocated plan needs memory for only 8 * 125,000 = 1,000,000 rows at any one time. Collocated Hash Join Performance The collated join plan has disappointing performance in this case. The query runs for around 25,300ms despite the same IO statistics as usual. This is much the worst result so far, so what went wrong? It turns out that cardinality estimation for the single partition scans of table T1 is slightly low. The properties of the Clustered Index Scan of T1 (graphic immediately above) show the estimation was for 121,951 rows. This is a small shortfall compared with the 125,000 rows actually encountered, but it was enough to cause the hash join to spill to physical tempdb: A level 1 spill doesn’t sound too bad, until you realize that the spill to tempdb probably occurs for each of the 41 partitions. As a side note, the cardinality estimation error is a little surprising because the system tables accurately show there are 125,000 rows in every partition of T1. Unfortunately, the optimizer uses regular column and index statistics to derive cardinality estimates here rather than system table information (e.g. sys.partitions). Collocated Merge Join We will never know how well the collocated parallel hash join plan might have worked without the cardinality estimation error (and the resulting 41 spills to tempdb) but we do know: Merge join does not require a memory grant; and Merge join was the optimizer’s preferred join option for a single partition join Putting this all together, what we would really like to see is the same collocated join strategy, but using merge join instead of hash join. Unfortunately, the current query optimizer cannot produce a collocated merge join; it only knows how to do collocated hash join. So where does this leave us? CROSS APPLY sys.partitions We can try to write our own collocated join query. We can use sys.partitions to find the partition numbers, and CROSS APPLY to get a count per partition, with a final step to sum the partial counts. The following query implements this idea: SELECT row_count = SUM(Subtotals.cnt) FROM ( -- Partition numbers SELECT p.partition_number FROM sys.partitions AS p WHERE p.[object_id] = OBJECT_ID(N'T1', N'U') AND p.index_id = 1 ) AS P CROSS APPLY ( -- Count per collocated join SELECT cnt = COUNT_BIG(*) FROM dbo.T1 AS T1 JOIN dbo.T2 AS T2 ON T2.TID = T1.TID WHERE $PARTITION.PFT(T1.TID) = p.partition_number AND $PARTITION.PFT(T2.TID) = p.partition_number ) AS SubTotals; The estimated plan is: The cardinality estimates aren’t all that good here, especially the estimate for the scan of the system table underlying the sys.partitions view. Nevertheless, the plan shape is heading toward where we would like to be. Each partition number from the system table results in a per-partition scan of T1 and T2, a one-to-many Merge Join, and a Stream Aggregate to compute the partial counts. The final Stream Aggregate just sums the partial counts. Execution time for this query is around 3,500ms, with the same IO statistics as always. This compares favourably with 5,000ms for the serial plan produced by the optimizer with the OPTION (MERGE JOIN) hint. This is another case of the sum of the parts being less than the whole – summing 41 partial counts from 41 single-partition merge joins is faster than a single merge join and count over all partitions. Even so, this single-threaded collocated merge join is not as quick as the original parallel hash join plan, which executed in 2,600ms. On the positive side, our collocated merge join uses only one logical processor and requires no memory grant. The parallel hash join plan used 16 threads and reserved 569 MB of memory:   Using a Temporary Table Our collocated merge join plan should benefit from parallelism. The reason parallelism is not being used is that the query references a system table. We can work around that by writing the partition numbers to a temporary table (or table variable): SET STATISTICS IO ON; DECLARE @s datetime2 = SYSUTCDATETIME();   CREATE TABLE #P ( partition_number integer PRIMARY KEY);   INSERT #P (partition_number) SELECT p.partition_number FROM sys.partitions AS p WHERE p.[object_id] = OBJECT_ID(N'T1', N'U') AND p.index_id = 1;   SELECT row_count = SUM(Subtotals.cnt) FROM #P AS p CROSS APPLY ( SELECT cnt = COUNT_BIG(*) FROM dbo.T1 AS T1 JOIN dbo.T2 AS T2 ON T2.TID = T1.TID WHERE $PARTITION.PFT(T1.TID) = p.partition_number AND $PARTITION.PFT(T2.TID) = p.partition_number ) AS SubTotals;   DROP TABLE #P;   SELECT DATEDIFF(Millisecond, @s, SYSUTCDATETIME()); SET STATISTICS IO OFF; Using the temporary table adds a few logical reads, but the overall execution time is still around 3500ms, indistinguishable from the same query without the temporary table. The problem is that the query optimizer still doesn’t choose a parallel plan for this query, though the removal of the system table reference means that it could if it chose to: In fact the optimizer did enter the parallel plan phase of query optimization (running search 1 for a second time): Unfortunately, the parallel plan found seemed to be more expensive than the serial plan. This is a crazy result, caused by the optimizer’s cost model not reducing operator CPU costs on the inner side of a nested loops join. Don’t get me started on that, we’ll be here all night. In this plan, everything expensive happens on the inner side of a nested loops join. Without a CPU cost reduction to compensate for the added cost of exchange operators, candidate parallel plans always look more expensive to the optimizer than the equivalent serial plan. Parallel Collocated Merge Join We can produce the desired parallel plan using trace flag 8649 again: SELECT row_count = SUM(Subtotals.cnt) FROM #P AS p CROSS APPLY ( SELECT cnt = COUNT_BIG(*) FROM dbo.T1 AS T1 JOIN dbo.T2 AS T2 ON T2.TID = T1.TID WHERE $PARTITION.PFT(T1.TID) = p.partition_number AND $PARTITION.PFT(T2.TID) = p.partition_number ) AS SubTotals OPTION (QUERYTRACEON 8649); The actual execution plan is: One difference between this plan and the collocated hash join plan is that a Repartition Streams exchange operator is used instead of Distribute Streams. The effect is similar, though not quite identical. The Repartition uses round-robin partitioning, meaning the next partition id is pushed to the next thread in sequence. The Distribute Streams exchange seen earlier used Demand partitioning, meaning the next partition id is pulled across the exchange by the next thread that is ready for more work. There are subtle performance implications for each partitioning option, but going into that would again take us too far off the main point of this post. Performance The important thing is the performance of this parallel collocated merge join – just 1350ms on a typical run. The list below shows all the alternatives from this post (all timings include creation, population, and deletion of the temporary table where appropriate) from quickest to slowest: Collocated parallel merge join: 1350ms Parallel hash join: 2600ms Collocated serial merge join: 3500ms Serial merge join: 5000ms Parallel merge join: 8400ms Collated parallel hash join: 25,300ms (hash spill per partition) The parallel collocated merge join requires no memory grant (aside from a paltry 1.2MB used for exchange buffers). This plan uses 16 threads at DOP 8; but 8 of those are (rather pointlessly) allocated to the parallel scan of the temporary table. These are minor concerns, but it turns out there is a way to address them if it bothers you. Parallel Collocated Merge Join with Demand Partitioning This final tweak replaces the temporary table with a hard-coded list of partition ids (dynamic SQL could be used to generate this query from sys.partitions): SELECT row_count = SUM(Subtotals.cnt) FROM ( VALUES (1),(2),(3),(4),(5),(6),(7),(8),(9),(10), (11),(12),(13),(14),(15),(16),(17),(18),(19),(20), (21),(22),(23),(24),(25),(26),(27),(28),(29),(30), (31),(32),(33),(34),(35),(36),(37),(38),(39),(40),(41) ) AS P (partition_number) CROSS APPLY ( SELECT cnt = COUNT_BIG(*) FROM dbo.T1 AS T1 JOIN dbo.T2 AS T2 ON T2.TID = T1.TID WHERE $PARTITION.PFT(T1.TID) = p.partition_number AND $PARTITION.PFT(T2.TID) = p.partition_number ) AS SubTotals OPTION (QUERYTRACEON 8649); The actual execution plan is: The parallel collocated hash join plan is reproduced below for comparison: The manual rewrite has another advantage that has not been mentioned so far: the partial counts (per partition) can be computed earlier than the partial counts (per thread) in the optimizer’s collocated join plan. The earlier aggregation is performed by the extra Stream Aggregate under the nested loops join. The performance of the parallel collocated merge join is unchanged at around 1350ms. Final Words It is a shame that the current query optimizer does not consider a collocated merge join (Connect item closed as Won’t Fix). The example used in this post showed an improvement in execution time from 2600ms to 1350ms using a modestly-sized data set and limited parallelism. In addition, the memory requirement for the query was almost completely eliminated  – down from 569MB to 1.2MB. The problem with the parallel hash join selected by the optimizer is that it attempts to process the full data set all at once (albeit using eight threads). It requires a large memory grant to hold all 5 million rows from table T1 across the eight hash tables, and does not take advantage of the divide-and-conquer opportunity offered by the common partitioning. The great thing about the collocated join strategies is that each parallel thread works on a single partition from both tables, reading rows, performing the join, and computing a per-partition subtotal, before moving on to a new partition. From a thread’s point of view… If you have trouble visualizing what is happening from just looking at the parallel collocated merge join execution plan, let’s look at it again, but from the point of view of just one thread operating between the two Parallelism (exchange) operators. Our thread picks up a single partition id from the Distribute Streams exchange, and starts a merge join using ordered rows from partition 1 of table T1 and partition 1 of table T2. By definition, this is all happening on a single thread. As rows join, they are added to a (per-partition) count in the Stream Aggregate immediately above the Merge Join. Eventually, either T1 (partition 1) or T2 (partition 1) runs out of rows and the merge join stops. The per-partition count from the aggregate passes on through the Nested Loops join to another Stream Aggregate, which is maintaining a per-thread subtotal. Our same thread now picks up a new partition id from the exchange (say it gets id 9 this time). The count in the per-partition aggregate is reset to zero, and the processing of partition 9 of both tables proceeds just as it did for partition 1, and on the same thread. Each thread picks up a single partition id and processes all the data for that partition, completely independently from other threads working on other partitions. One thread might eventually process partitions (1, 9, 17, 25, 33, 41) while another is concurrently processing partitions (2, 10, 18, 26, 34) and so on for the other six threads at DOP 8. The point is that all 8 threads can execute independently and concurrently, continuing to process new partitions until the wider job (of which the thread has no knowledge!) is done. This divide-and-conquer technique can be much more efficient than simply splitting the entire workload across eight threads all at once. Related Reading Understanding and Using Parallelism in SQL Server Parallel Execution Plans Suck © 2013 Paul White – All Rights Reserved Twitter: @SQL_Kiwi

    Read the article

  • Passthrough Objects – Duck Typing++

    - by EltonStoneman
    [Source: http://geekswithblogs.net/EltonStoneman] Can't see a genuine use for this, but I got the idea in my head and wanted to work it through. It's an extension to the idea of duck typing, for scenarios where types have similar behaviour, but implemented in differently-named members. So you may have a set of objects you want to treat as an interface, which don't implement the interface explicitly, and don't have the same member names so they can't be duck-typed into implicitly implementing the interface. In a fictitious example, I want to call Get on whichever ICache implementation is current, and have the call passed through to the relevant method – whether it's called Read, Retrieve or whatever: A sample implementation is up on github here: PassthroughSample. This uses Castle's DynamicProxy behind the scenes in the same way as my duck typing sample, but allows you to configure the passthrough to specify how the inner (implementation) and outer (interface) members are mapped:       var setup = new Passthrough();     var cache = setup.Create("PassthroughSample.Tests.Stubs.AspNetCache, PassthroughSample.Tests")                             .WithPassthrough("Name", "CacheName")                             .WithPassthrough("Get", "Retrieve")                             .WithPassthrough("Set", "Insert")                             .As<ICache>(); - or using some ugly Lambdas to avoid the strings :     Expression<Func<ICache, string, object>> get = (o, s) => o.Get(s);     Expression<Func<Memcached, string, object>> read = (i, s) => i.Read(s);     Expression<Action<ICache, string, object>> set = (o, s, obj) => o.Set(s, obj);     Expression<Action<Memcached, string, object>> insert = (i, s, obj) => i.Put(s, obj);       ICache cache = new Passthrough<ICache, Memcached>()                     .Create()                     .WithPassthrough(o => o.Name, i => i.InstanceName)                     .WithPassthrough(get, read)                     .WithPassthrough(set, insert)                     .As();   - or even in config:   ICache cache = Passthrough.GetConfigured<ICache>(); ...  <passthrough>     <types>       <typename="PassthroughSample.Tests.Stubs.ICache, PassthroughSample.Tests"             passesThroughTo="PassthroughSample.Tests.Stubs.AppFabricCache, PassthroughSample.Tests">         <members>           <membername="Name"passesThroughTo="RegionName"/>           <membername="Get"passesThroughTo="Out"/>           <membername="Set"passesThroughTo="In"/>         </members>       </type>   Possibly useful for injecting stubs for dependencies in tests, when your application code isn't using an IoC container. Possibly it also has an alternative implementation using .NET 4.0 dynamic objects, rather than the dynamic proxy.

    Read the article

  • Backup Meta-Data

    - by BuckWoody
    I'm working on a PowerShell script to show me the trending durations of my backup activities. The first thing I need is the data, so I looked at the Standard Reports in SQL Server Management Studio, and found a report that suited my needs, so I pulled out the script that it runs and modified it to this T-SQL Script. A few words here - you need to be in the MSDB database for this to run, and you can add a WHERE clause to limit to a database, timeframe, type of backup, whatever. For that matter, I won't use all of the data in this query in my PowerShell script, but it gives me lots of avenues to graph: SELECT distinct t1.name AS 'DatabaseName' ,(datediff( ss,  t3.backup_start_date, t3.backup_finish_date)) AS 'DurationInSeconds' ,t3.user_name AS 'UserResponsible' ,t3.name AS backup_name ,t3.description ,t3.backup_start_date ,t3.backup_finish_date ,CASE WHEN t3.type = 'D' THEN 'Database' WHEN t3.type = 'L' THEN 'Log' WHEN t3.type = 'F' THEN 'FileOrFilegroup' WHEN t3.type = 'G' THEN 'DifferentialFile' WHEN t3.type = 'P' THEN 'Partial' WHEN t3.type = 'Q' THEN 'DifferentialPartial' END AS 'BackupType' ,t3.backup_size AS 'BackupSizeKB' ,t6.physical_device_name ,CASE WHEN t6.device_type = 2 THEN 'Disk' WHEN t6.device_type = 102 THEN 'Disk' WHEN t6.device_type = 5 THEN 'Tape' WHEN t6.device_type = 105 THEN 'Tape' END AS 'DeviceType' ,t3.recovery_model  FROM sys.databases t1 INNER JOIN backupset t3 ON (t3.database_name = t1.name )  LEFT OUTER JOIN backupmediaset t5 ON ( t3.media_set_id = t5.media_set_id ) LEFT OUTER JOIN backupmediafamily t6 ON ( t6.media_set_id = t5.media_set_id ) ORDER BY backup_start_date DESC I'll munge this into my Excel PowerShell chart script tomorrow. Script Disclaimer, for people who need to be told this sort of thing: Never trust any script, including those that you find here, until you understand exactly what it does and how it will act on your systems. Always check the script on a test system or Virtual Machine, not a production system. Yes, there are always multiple ways to do things, and this script may not work in every situation, for everything. It’s just a script, people. All scripts on this site are performed by a professional stunt driver on a closed course. Your mileage may vary. Void where prohibited. Offer good for a limited time only. Keep out of reach of small children. Do not operate heavy machinery while using this script. If you experience blurry vision, indigestion or diarrhea during the operation of this script, see a physician immediately. Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!

    Read the article

  • SQL SERVER – guest User and MSDB Database – Enable guest User on MSDB Database

    - by pinaldave
    I have written a few articles recently on the subject of guest account. Here’s a quick list of these articles: SQL SERVER – Disable Guest Account – Serious Security Issue SQL SERVER – Force Removing User from Database – Fix: Error: Could not drop login ‘test’ as the user is currently logged in. SQL SERVER – Detecting guest User Permissions – guest User Access Status One of the advices which I gave in all the three blog posts was: Disable the guest user in the user-created database. Additionally, I have mentioned that one should let the user account become enabled in MSDB database. I got many questions asking if there is any specific reason why this should be kept enabled, questions like, “What is the reason that MSDB database needs guest user?” Honestly, I did not know that the concept of the guest user will create so much interest in the readers. So now let’s turn this blog post into questions and answers format. Q: What will happen if the guest user is disabled in MSDB database? A:  Lots of bad things will happen. Error 916 - Logins can connect to this instance of SQL Server but they do not have specific permissions in a database to receive the permissions of the guest user. Q: How can I determine if the guest user is enabled or disabled for any specific database? A: There are many ways to do this. Make sure that you run each of these methods with the context of the database. For an example for msdb database, you can run the following code: USE msdb; SELECT name, permission_name, state_desc FROM sys.database_principals dp INNER JOIN sys.server_permissions sp ON dp.principal_id = sp.grantee_principal_id WHERE name = 'guest' AND permission_name = 'CONNECT' There are many other methods to detect the guest user status. Read them here: Detecting guest User Permissions – guest User Access Status Q: What is the default status of the guest user account in database? A: Enabled in master, TempDb, and MSDB. Disabled in model database. Q: Why is the default status of the guest user disabled in model database? A: It is not recommended to enable the guest in user database as it can introduce serious security threat. It can seriously damage the database if configured incorrectly. Read more here: Disable Guest Account – Serious Security Issue Q: How to disable guest user? A: REVOKE CONNECT FROM guest Q: How to enable guest user? A: GRANT CONNECT TO guest Did I miss any critical question in the list? Please leave your question as a comment and I will add it to this list. Reference: Pinal Dave (http://blog.sqlauthority.com) Filed under: PostADay, SQL, SQL Authority, SQL Query, SQL Security, SQL Server, SQL Tips and Tricks, T SQL, Technology

    Read the article

  • How to get over “Did I lock the door?” syndrome

    - by Boonei
    I am person who always asks myself  ”Did I lock the house door?”,  And I do ask that question when I have almost reached office. I don’t have a bad memory or I am not a “forget it all after a min person”. Infact I have a fantastic memory of things. This problem has been haunting me for a very long time. My wife used to always have a angry face after we had get down from the car. Because after we have walked for about 20 yards I would run back to the car to check if I had locked the car, you see this problem exists for all locked objects. This happens everyday all round the year. Now a days I don’t have the problem ! I did not get the solution from any doctor or any book that that talks about my inner mind. It was a practical advice given by my aunt….. When I told her that I had this problem, she smiled and said its very very easy to get around this. I was stunned. The solution she gave me was simple. After I had locked the door, should hold the lock and look at it for 5 sec and say to myself   “I have locked the door”. Believe me it works like a charm. The reason why it works is my aunt goes to explain, that your mind always thinks twice of important things that we do on our daily life and raises doubts after sometime. The only way to stop is it by looking at it, holding it and telling yourself that its ok and its done. This holds good for all the things that you generally doubt like, did I turn off the AC?, did I turn off the lights in the house when I left?. Just look at it for 5 sec, hold it tell yourself its done. You will not look back. Image credit [Håkan Dahlström]   This article titled,How to get over “Did I lock the door?” syndrome, was originally published at Tech Dreams. Grab our rss feed or fan us on Facebook to get updates from us.

    Read the article

  • Backup Meta-Data

    - by BuckWoody
    I'm working on a PowerShell script to show me the trending durations of my backup activities. The first thing I need is the data, so I looked at the Standard Reports in SQL Server Management Studio, and found a report that suited my needs, so I pulled out the script that it runs and modified it to this T-SQL Script. A few words here - you need to be in the MSDB database for this to run, and you can add a WHERE clause to limit to a database, timeframe, type of backup, whatever. For that matter, I won't use all of the data in this query in my PowerShell script, but it gives me lots of avenues to graph: SELECT distinct t1.name AS 'DatabaseName' ,(datediff( ss,  t3.backup_start_date, t3.backup_finish_date)) AS 'DurationInSeconds' ,t3.user_name AS 'UserResponsible' ,t3.name AS backup_name ,t3.description ,t3.backup_start_date ,t3.backup_finish_date ,CASE WHEN t3.type = 'D' THEN 'Database' WHEN t3.type = 'L' THEN 'Log' WHEN t3.type = 'F' THEN 'FileOrFilegroup' WHEN t3.type = 'G' THEN 'DifferentialFile' WHEN t3.type = 'P' THEN 'Partial' WHEN t3.type = 'Q' THEN 'DifferentialPartial' END AS 'BackupType' ,t3.backup_size AS 'BackupSizeKB' ,t6.physical_device_name ,CASE WHEN t6.device_type = 2 THEN 'Disk' WHEN t6.device_type = 102 THEN 'Disk' WHEN t6.device_type = 5 THEN 'Tape' WHEN t6.device_type = 105 THEN 'Tape' END AS 'DeviceType' ,t3.recovery_model  FROM sys.databases t1 INNER JOIN backupset t3 ON (t3.database_name = t1.name )  LEFT OUTER JOIN backupmediaset t5 ON ( t3.media_set_id = t5.media_set_id ) LEFT OUTER JOIN backupmediafamily t6 ON ( t6.media_set_id = t5.media_set_id ) ORDER BY backup_start_date DESC I'll munge this into my Excel PowerShell chart script tomorrow. Script Disclaimer, for people who need to be told this sort of thing: Never trust any script, including those that you find here, until you understand exactly what it does and how it will act on your systems. Always check the script on a test system or Virtual Machine, not a production system. Yes, there are always multiple ways to do things, and this script may not work in every situation, for everything. It’s just a script, people. All scripts on this site are performed by a professional stunt driver on a closed course. Your mileage may vary. Void where prohibited. Offer good for a limited time only. Keep out of reach of small children. Do not operate heavy machinery while using this script. If you experience blurry vision, indigestion or diarrhea during the operation of this script, see a physician immediately. Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!

    Read the article

  • Is your test method self-validating ?

    - by mehfuzh
    Writing state of art unit tests that can validate your every part of the framework is challenging and interesting at the same time, its like becoming a samurai. One of the key concept in this is to keep our test synced all the time as underlying code changes and thus breaking them to the furthest unit as possible.  This also means, we should avoid  multiple conditions embedded in a single test. Let’s consider the following example of transfer funds. [Fact] public void ShouldAssertTranserFunds() {     var currencyService = Mock.Create<ICurrencyService>();     //// current rate     Mock.Arrange(() => currencyService.GetConversionRate("AUS", "CAD")).Returns(0.88f);       Account to = new Account { Currency = "AUS", Balance = 120 };     Account from = new Account { Currency = "CAD" };       AccountService accService = new AccountService(currencyService);       Assert.Throws<InvalidOperationException>(() => accService.TranferFunds(to, from, 200f));       accService.TranferFunds(to, from, 100f);       Assert.Equal(from.Balance, 88);     Assert.Equal(20, to.Balance); } At first look,  it seems ok but as you look more closely , it is actually doing two tasks in one test. At line# 10 it is trying to validate the exception for invalid fund transfer and finally it is asserting if the currency conversion is successfully made. Here, the name of the test itself is pretty vague. The first rule for writing unit test should always reflect to inner working of the target code, where just by looking at their names it is self explanatory. Having a obscure name for a test method not only increase the chances of cluttering the test code, but it also gives the opportunity to add multiple paths into it and eventually makes things messy as possible. I would rater have two test methods that explicitly describes its intent and are more self-validating. ShouldThrowExceptionForInvalidTransferOperation ShouldAssertTransferForExpectedConversionRate Having, this type of breakdown also helps us pin-point reported bugs easily rather wasting any time on debugging for something more general and can minimize confusion among team members. Finally, we should always make our test F.I.R.S.T ( Fast.Independent.Repeatable.Self-validating.Timely) [ Bob martin – Clean Code]. Only this will be enough to ensure, our test is as simple and clean as possible.   Hope that helps

    Read the article

  • Working with Joins in LINQ

    - by vik20000in
    While working with data most of the time we have to work with relation between different lists of data. Many a times we want to fetch data from both the list at once. This requires us to make different kind of joins between the lists of data. LINQ support different kinds of join Inner Join     List<Customer> customers = GetCustomerList();     List<Supplier> suppliers = GetSupplierList();      var custSupJoin =         from sup in suppliers         join cust in customers on sup.Country equals cust.Country         select new { Country = sup.Country, SupplierName = sup.SupplierName, CustomerName = cust.CompanyName }; Group Join – where By the joined dataset is also grouped.     List<Customer> customers = GetCustomerList();     List<Supplier> suppliers = GetSupplierList();      var custSupQuery =         from sup in suppliers         join cust in customers on sup.Country equals cust.Country into cs         select new { Key = sup.Country, Items = cs }; We can also work with the Left outer join in LINQ like this.     List<Customer> customers = GetCustomerList();     List<Supplier> suppliers = GetSupplierList();      var supplierCusts =         from sup in suppliers         join cust in customers on sup.Country equals cust.Country into cs         from c in cs.DefaultIfEmpty()  // DefaultIfEmpty preserves left-hand elements that have no matches on the right side         orderby sup.SupplierName         select new { Country = sup.Country, CompanyName = c == null ? "(No customers)" : c.CompanyName,                      SupplierName = sup.SupplierName};Vikram

    Read the article

< Previous Page | 63 64 65 66 67 68 69 70 71 72 73 74  | Next Page >