I need to analyze a task that starts with the code below but I couldn't figure out what the LINQ part is doing. Any leads are appreciated
foreach (var item in list.GroupBy(x => x.AccountNumber).Select(g => g.First()))
{
...
}
Some roughly-equivalent code (i.e. has the same function, but works slightly differently) would be:
var seenAccountNumbers = new HashSet<int>(); // Or some other data type?
foreach (var item in list)
{
if (seenAccountNumbers.Add(item.AccountNumber))
{
...
}
}
This code is a (somewhat wasteful) way of getting the first item by account number. It's wasteful because there's no reason to group everything before trying to find the first item per group.
The same thing can be implemented with an iterator function by iterating over all items in the input list and keeping track of all the AccountNumber values found so far. When a new one is found, yield it and add it to the tracking list. Or rather, HashSet.
In fact, that's how MoreLinq's DistinctBy operator is implemented :
var knownKeys = new HashSet<TKey>(comparer);
foreach (var element in source)
{
if (knownKeys.Add(keySelector(element)))
yield return element;
}
From the method's description:
Returns all distinct elements of the given source, where "distinctness"is determined via a projection and the default equality comparer for the projected type.
If a key is seen multiple times, only the first element with that key is returned.
The question's code can be replaced with :
foreach (var item in list.DistinctBy(x => x.AccountNumber))
{...
}
Create a dictionary, with the AccountNumber as Key, and put all your items from list, in that dictionary. That is about what happens.
You will overwrite items, with the same key, and a randomly last element, will stay in the dictionary. There is no order ensured when using GroupBy, so it doesn't matter if you choose First or Last element at the end, it just has the meaning of "pick one" (random).
var dict = new Dictionary<KeyType, ElementType>();
foreach(var item in list)
if (!dict.ContainsKey(item.AccountNumber))
dict[item.AccountNumber] = item;
You original iteration would now be
foreach(var item in dict.Values)
{
.....
}
To ask for Non-LINQ solution is not so strange, cause LINQ offers never the most performant solution, it's just short writing and fast coding.
I have a list of Items, each containing a field of Type integer.
I want to filter my list to get only the items that match a given list of integers.
The code I have now works but I know it could be optimized.
Class Item
{
int ID;
//Other fields & methods that are irrelevant here
}
//Selection method
IEnumerable<Item> SelectItems(List<Item> allItems, List<int> toSelect)
{
return allItems.Where(x => toSelect.Contains(x.ID));
}
The problem I have is that I iterate through allItems and in each iteration I iterate through toSelect.
I have the feeling it is possible to be much more effective but I don't know how I can achieve this with Linq.
This might also be an already asked question as I don't know how this is called in English. This feels kind of stupid because I don't know how to formulate this properly in a seach engine.
You can use Join which is more efficient because it's using a set based approach:
var selectedItems = from item in allItems
join id in toSelect
on item.Id equals id
select item;
return selectedItems;
Another way which is more efficient is to use a HashSet<int> instead of a list:
IEnumerable<Item> SelectItems(List<Item> allItems, HashSet<int> toSelect)
{
return allItems.Where(x => toSelect.Contains(x.ID));
}
There are two ways to approach this.
Currently you have O(N×M) performance (where N is the size of allItems and M is the size of toSelect).
If you're just trying to reduce it easily, then you could reduce it to O(N)+O(M) by creating a hash-set of toSelect:
var matches = new HashSet<int>(toSelect);
return allItems.Where(x => matches.Contains(x.ID));
However, this is still going to be dominated by N - the size of allItems.
A better long term approach may be to pre-index the data (and keep it indexed) by Id. So instead of allItems being a List<T> - it could be a Dictionary<int, T>. Note that building the dictionary can be expensive, so you don't want to do this every time you want to search : the key is to do this once at the start (and keep it maintained). Then this becomes O(M) (the size of toSelect, which is usually small), since dictionary lookups are O(1).
IEnumerable<Item> SelectItems(Dictionary<int, Item> allItems, List<int> toSelect)
{
foreach(var id in toSelect)
{
if (allItems.TryGetValue(id, out var found))
yield return found;
}
}
(there is no need to pre-hash toSelect since we aren't checking it for Contains)
Consider you have two lists in C#, first list contains elements of TypeOne and second list contains elements of TypeTwo:
TypeOne
{
int foo;
int bar;
}
TypeTwo
{
int baz;
int qux;
}
Now I need to find elements ( with some property value ) in the first list that don't exist in the second list, and similarly I want to find elements in the second list that don't exist in the first list. (There are only zero or one occurences in either lists.)
What I tried so far is to iterate both lists like this:
foreach (var item in firstList)
{
if (!secondList.Any(a=> a.baz == item.foo)
{
// Item is in the first list but not in second list.
}
}
and again:
foreach (var item in secondList)
{
if (!firstList.Any(a=> a.foo == item.baz)
{
// Item is in the second list but not in first list.
}
}
I hardly think this is a good way to do what I want. I'm iterating my lists two times and use Any in each of them which also iterates the list. So too many iterations.
What is the most efficient way to achieve this?
I am afraid there is no prebuild solution for this, so the best we can do is optimize as much as possible. We only have to iterate the first list, because everything that is in second will be compared already
// First we need copies to operate on
var firstCopy = new List<TypeOne>(firstList);
var secondCopy = new List<TypeTwo>(secondList);
// Now we iterate the first list once complete
foreach (var typeOne in firstList)
{
var match = secondCopy.FirstOrDefault(s => s.baz == typeOne.foo);
if (match == null)
{
// Item in first but not in second
}
else
{
// Match is duplicate and shall be removed from both
firstCopy.Remove(typeOne);
secondCopy.Remove(match);
}
}
After running this both copies will only contain the values which are unique in this instance. This not only reduces it to half the number of iterations but also constantly improves because the second copy shrinks with each match.
Use this LINQ Query.
var result1 = secondList.Where(p2 => !firstList.Any(p1 => p1.foo == p2.baz));
var result2=firstList.Where(p1=> !secondList.Any(p2=> p2.foo == p1.baz);
First, I know this isn't possible out of the box because of obvious reasons.
foreach(string item in myListOfStrings) {
myListOfStrings.Remove(item);
}
The snipped above is one of the most horrible things I've ever seen. So, how do you achieve it then? You could iterate through the list backwards using for, but I don't like this solution either.
What I'm wondering is: Is there a method/extensions that returns an IEnumerable from the current list, something like a floating copy? LINQ has numerous extension methods that do exactly this, but you always have to do something with it, such as filtering (where, take...).
I'm looking forward to something like this:
foreach(string item in myListOfStrings.Shadow()) {
myListOfStrings.Remove(item);
}
where as .Shadow() is:
public static IEnumerable<T> Shadow<T>(this IEnumerable<T> source) {
return new IEnumerable<T>(source);
// or return source.Copy()
// or return source.TakeAll();
}
Example
foreach(ResponseFlags flag in responseFlagsList.Shadow()) {
switch(flag) {
case ResponseFlags.Case1:
...
case ResponseFlags.Case2:
...
}
...
this.InvokeSomeVoidEvent(flag)
responseFlagsList.Remove(flag);
}
Solution
This is how I solved it, and it works like a charm:
public static IEnumerable<T> Shadow<T>(this IEnumerable<T> source) where T: new() {
foreach(T item in source)
yield return item;
}
It's not that super fast (obviously), but it's safe and exactly what I intended to do.
Removing multiple elements from a list 1 by 1 is a C# anti-pattern due to how lists are implemented.
Of course, it can be done with a for loop (instead of foreach). Or it can be done by making a copy of the list. But here is why it should not be done. On a list of 100000 random integers, this takes 2500 ms on my machine:
foreach (var x in listA.ToList())
if (x % 2 == 0)
listA.Remove(x);
and this takes 1250 ms:
for (int i = 0; i < listA.Count; i++)
if (listA[i] % 2 == 0)
listA.RemoveAt(i--);
while these two take 5 and 2 ms respectively:
listB = listB.Where(x => x % 2 != 0).ToList();
listB.RemoveAll(x => x % 2 == 0);
This is because when you remove an element from a list, you are actually deleting from an array, and this is O(N) time, as you need to shift each element after the deleted element one position to the left. On average, this will be N/2 elements.
Remove(element) also needs to find the element before removing it. So Remove(element) will actually always take N steps - elementindex steps to find the element, N - elementindex steps to remove it - in total, N steps.
RemoveAt(index) doesn't have to find the element, but it still has to shift the underlying array, so on average, a RemoveAt is N/2 steps.
The end result is O(N^2) complexity either way, as you're removing up to N elements.
Instead, you should use Linq, which will modify the entire list in O(N) time, or roll your own, but you should not use Remove (or RemoveAt) in a loop.
Why not just do:
foreach(string item in myListOfStrings.ToList())
{
myListOfStrings.Remove(item);
}
To create a copy of the original and use for iterating, then remove from the existing.
If you really need your extension method you could perhaps create something more readable to the user such as:
public static IEnumerable<T> Shadow<T>(this IEnumerable<T> items)
{
if (items == null)
throw new NullReferenceException("Items cannot be null");
List<T> list = new List<T>();
foreach (var item in items)
{
list.Add(item);
}
return list;
}
Which is essentially the same as .ToList().
Calling:
foreach(string item in myListOfStrings.Shadow())
You do not LINQ extension methods for this - you can create a new list explicitly, like this:
foreach(string item in new List<string>(myListOfStrings)) {
myListOfStrings.Remove(item);
}
You have to create a copy of the original list while iterating as below:
var myListOfStrings = new List<string>();
myListOfStrings.Add("1");
myListOfStrings.Add("2");
myListOfStrings.Add("3");
myListOfStrings.Add("4");
myListOfStrings.Add("5");
foreach (string item in myListOfStrings.ToList())
{
myListOfStrings.Remove(item);
}
Your example removes all items from the string, so it's equivalent to:
myListOfStrings.Clear();
It is also equivalent to:
myListOfStrings.RemoveAll(x => true); // Empties myListOfStrings
But what I think you're looking for is a way to remove items for which a predicate is true - which is what RemoveAll() does.
So you could write, for example:
myListOfStrings.RemoveAll(x => x == "TEST"); // Modifies myListOfStrings
Or use any other predicate.
However, that changes the ORIGINAL list; If you just want a copy of the list with certain items removed, you can just use normal Linq:
// Note != instead of == as used in Removeall(),
// because the logic here is reversed.
var filteredList = myListOfStrings.Where(x => x != "TEST").ToList();
Picking up on the answer of svinja I do believe the most efficient way of solving this problem is by doing:
for (int i = 0; i < listA.Count;) {
if (listA[i] % 2 == 0)
listA.RemoveAt(i);
else
i++;
}
It improves on the answer by removing unnecessary sums and subtractions.
What is the best way to approach removing items from a collection in C#, once the item is known, but not it's index. This is one way to do it, but it seems inelegant at best.
//Remove the existing role assignment for the user.
int cnt = 0;
int assToDelete = 0;
foreach (SPRoleAssignment spAssignment in workspace.RoleAssignments)
{
if (spAssignment.Member.Name == shortName)
{
assToDelete = cnt;
}
cnt++;
}
workspace.RoleAssignments.Remove(assToDelete);
What I would really like to do is find the item to remove by property (in this case, name) without looping through the entire collection and using 2 additional variables.
If RoleAssignments is a List<T> you can use the following code.
workSpace.RoleAssignments.RemoveAll(x =>x.Member.Name == shortName);
If you want to access members of the collection by one of their properties, you might consider using a Dictionary<T> or KeyedCollection<T> instead. This way you don't have to search for the item you're looking for.
Otherwise, you could at least do this:
foreach (SPRoleAssignment spAssignment in workspace.RoleAssignments)
{
if (spAssignment.Member.Name == shortName)
{
workspace.RoleAssignments.Remove(spAssignment);
break;
}
}
#smaclell asked why reverse iteration was more efficient in in a comment to #sambo99.
Sometimes it's more efficient. Consider you have a list of people, and you want to remove or filter all customers with a credit rating < 1000;
We have the following data
"Bob" 999
"Mary" 999
"Ted" 1000
If we were to iterate forward, we'd soon get into trouble
for( int idx = 0; idx < list.Count ; idx++ )
{
if( list[idx].Rating < 1000 )
{
list.RemoveAt(idx); // whoops!
}
}
At idx = 0 we remove Bob, which then shifts all remaining elements left. The next time through the loop idx = 1, but
list[1] is now Ted instead of Mary. We end up skipping Mary by mistake. We could use a while loop, and we could introduce more variables.
Or, we just reverse iterate:
for (int idx = list.Count-1; idx >= 0; idx--)
{
if (list[idx].Rating < 1000)
{
list.RemoveAt(idx);
}
}
All the indexes to the left of the removed item stay the same, so you don't skip any items.
The same principle applies if you're given a list of indexes to remove from an array. In order to keep things straight you need to sort the list and then remove the items from highest index to lowest.
Now you can just use Linq and declare what you're doing in a straightforward manner.
list.RemoveAll(o => o.Rating < 1000);
For this case of removing a single item, it's no more efficient iterating forwards or backwards. You could also use Linq for this.
int removeIndex = list.FindIndex(o => o.Name == "Ted");
if( removeIndex != -1 )
{
list.RemoveAt(removeIndex);
}
If it's an ICollection then you won't have a RemoveAll method. Here's an extension method that will do it:
public static void RemoveAll<T>(this ICollection<T> source,
Func<T, bool> predicate)
{
if (source == null)
throw new ArgumentNullException("source", "source is null.");
if (predicate == null)
throw new ArgumentNullException("predicate", "predicate is null.");
source.Where(predicate).ToList().ForEach(e => source.Remove(e));
}
Based on:
http://phejndorf.wordpress.com/2011/03/09/a-removeall-extension-for-the-collection-class/
For a simple List structure the most efficient way seems to be using the Predicate RemoveAll implementation.
Eg.
workSpace.RoleAssignments.RemoveAll(x =>x.Member.Name == shortName);
The reasons are:
The Predicate/Linq RemoveAll method is implemented in List and has access to the internal array storing the actual data. It will shift the data and resize the internal array.
The RemoveAt method implementation is quite slow, and will copy the entire underlying array of data into a new array. This means reverse iteration is useless for List
If you are stuck implementing this in a the pre c# 3.0 era. You have 2 options.
The easily maintainable option. Copy all the matching items into a new list and and swap the underlying list.
Eg.
List<int> list2 = new List<int>() ;
foreach (int i in GetList())
{
if (!(i % 2 == 0))
{
list2.Add(i);
}
}
list2 = list2;
Or
The tricky slightly faster option, which involves shifting all the data in the list down when it does not match and then resizing the array.
If you are removing stuff really frequently from a list, perhaps another structure like a HashTable (.net 1.1) or a Dictionary (.net 2.0) or a HashSet (.net 3.5) are better suited for this purpose.
What type is the collection? If it's List, you can use the helpful "RemoveAll":
int cnt = workspace.RoleAssignments
.RemoveAll(spa => spa.Member.Name == shortName)
(This works in .NET 2.0. Of course, if you don't have the newer compiler, you'll have to use "delegate (SPRoleAssignment spa) { return spa.Member.Name == shortName; }" instead of the nice lambda syntax.)
Another approach if it's not a List, but still an ICollection:
var toRemove = workspace.RoleAssignments
.FirstOrDefault(spa => spa.Member.Name == shortName)
if (toRemove != null) workspace.RoleAssignments.Remove(toRemove);
This requires the Enumerable extension methods. (You can copy the Mono ones in, if you are stuck on .NET 2.0). If it's some custom collection that cannot take an item, but MUST take an index, some of the other Enumerable methods, such as Select, pass in the integer index for you.
This is my generic solution
public static IEnumerable<T> Remove<T>(this IEnumerable<T> items, Func<T, bool> match)
{
var list = items.ToList();
for (int idx = 0; idx < list.Count(); idx++)
{
if (match(list[idx]))
{
list.RemoveAt(idx);
idx--; // the list is 1 item shorter
}
}
return list.AsEnumerable();
}
It would look much simpler if extension methods support passing by reference !
usage:
var result = string[]{"mike", "john", "ali"}
result = result.Remove(x => x.Username == "mike").ToArray();
Assert.IsTrue(result.Length == 2);
EDIT: ensured that the list looping remains valid even when deleting items by decrementing the index (idx).
Here is a pretty good way to do it
http://support.microsoft.com/kb/555972
System.Collections.ArrayList arr = new System.Collections.ArrayList();
arr.Add("1");
arr.Add("2");
arr.Add("3");
/*This throws an exception
foreach (string s in arr)
{
arr.Remove(s);
}
*/
//where as this works correctly
Console.WriteLine(arr.Count);
foreach (string s in new System.Collections.ArrayList(arr))
{
arr.Remove(s);
}
Console.WriteLine(arr.Count);
Console.ReadKey();
There is another approach you can take depending on how you're using your collection. If you're downloading the assignments one time (e.g., when the app runs), you could translate the collection on the fly into a hashtable where:
shortname => SPRoleAssignment
If you do this, then when you want to remove an item by short name, all you need to do is remove the item from the hashtable by key.
Unfortunately, if you're loading these SPRoleAssignments a lot, that obviously isn't going to be any more cost efficient in terms of time. The suggestions other people made about using Linq would be good if you're using a new version of the .NET Framework, but otherwise, you'll have to stick to the method you're using.
Similar to Dictionary Collection point of view, I have done this.
Dictionary<string, bool> sourceDict = new Dictionary<string, bool>();
sourceDict.Add("Sai", true);
sourceDict.Add("Sri", false);
sourceDict.Add("SaiSri", true);
sourceDict.Add("SaiSriMahi", true);
var itemsToDelete = sourceDict.Where(DictItem => DictItem.Value == false);
foreach (var item in itemsToDelete)
{
sourceDict.Remove(item.Key);
}
Note:
Above code will fail in .Net Client Profile (3.5 and 4.5) also some viewers mentioned it is
Failing for them in .Net4.0 as well not sure which settings are causing the problem.
So replace with below code (.ToList()) for Where statement, to avoid that error. “Collection was modified; enumeration operation may not execute.”
var itemsToDelete = sourceDict.Where(DictItem => DictItem.Value == false).ToList();
Per MSDN From .Net4.5 onwards Client Profile are discontinued. http://msdn.microsoft.com/en-us/library/cc656912(v=vs.110).aspx
Save your items first, than delete them.
var itemsToDelete = Items.Where(x => !!!your condition!!!).ToArray();
for (int i = 0; i < itemsToDelete.Length; ++i)
Items.Remove(itemsToDelete[i]);
You need to override GetHashCode() in your Item class.
The best way to do it is by using linq.
Example class:
public class Product
{
public string Name { get; set; }
public string Price { get; set; }
}
Linq query:
var subCollection = collection1.RemoveAll(w => collection2.Any(q => q.Name == w.Name));
This query will remove all elements from collection1 if Name match any element Name from collection2
Remember to use: using System.Linq;
To do this while looping through the collection and not to get the modifying a collection exception, this is the approach I've taken in the past (note the .ToList() at the end of the original collection, this creates another collection in memory, then you can modify the existing collection)
foreach (SPRoleAssignment spAssignment in workspace.RoleAssignments.ToList())
{
if (spAssignment.Member.Name == shortName)
{
workspace.RoleAssignments.Remove(spAssignment);
}
}
If you have got a List<T>, then List<T>.RemoveAll is your best bet. There can't be anything more efficient. Internally it does the array moving in one shot, not to mention it is O(N).
If all you got is an IList<T> or an ICollection<T> you got roughly these three options:
public static void RemoveAll<T>(this IList<T> ilist, Predicate<T> predicate) // O(N^2)
{
for (var index = ilist.Count - 1; index >= 0; index--)
{
var item = ilist[index];
if (predicate(item))
{
ilist.RemoveAt(index);
}
}
}
or
public static void RemoveAll<T>(this ICollection<T> icollection, Predicate<T> predicate) // O(N)
{
var nonMatchingItems = new List<T>();
// Move all the items that do not match to another collection.
foreach (var item in icollection)
{
if (!predicate(item))
{
nonMatchingItems.Add(item);
}
}
// Clear the collection and then copy back the non-matched items.
icollection.Clear();
foreach (var item in nonMatchingItems)
{
icollection.Add(item);
}
}
or
public static void RemoveAll<T>(this ICollection<T> icollection, Func<T, bool> predicate) // O(N^2)
{
foreach (var item in icollection.Where(predicate).ToList())
{
icollection.Remove(item);
}
}
Go for either 1 or 2.
1 is lighter on memory and faster if you have less deletes to perform (i.e. predicate is false most of the times).
2 is faster if you have more deletes to perform.
3 is the cleanest code but performs poorly IMO. Again all that depends on input data.
For some benchmarking details see https://github.com/dotnet/BenchmarkDotNet/issues/1505
A lot of good responses here; I especially like the lambda expressions...very clean. I was remiss, however, in not specifying the type of Collection. This is a SPRoleAssignmentCollection (from MOSS) that only has Remove(int) and Remove(SPPrincipal), not the handy RemoveAll(). So, I have settled on this, unless there is a better suggestion.
foreach (SPRoleAssignment spAssignment in workspace.RoleAssignments)
{
if (spAssignment.Member.Name != shortName) continue;
workspace.RoleAssignments.Remove((SPPrincipal)spAssignment.Member);
break;
}