|
I have some image preview control based on a Panel which remembers current preview image position. Its OnPaint method does nothing except drawing viewed part of the image using this DrawImage override:
Graphics.DrawImage(Image, Rectangle, Rectangle, GraphicsUnit)
This draws only desired part of an image. The preview control is very small, so the whole painting should be fast, but it isn't.
Drawing 50x50 px rectangular part from one image to second - for some unknown reason - depends on the size of the source image  When this method draws little rectangle from 7 Mpx image, it tooks about 200 ms - that's not real-time, when image is moved and the control is repeatedly invalidated.
I saw similar preview controls, where the scrolling of image were smooth and doesn't depend on image size. These controls were part of the MFC apps. Then I tried professional .NET component for image previews, which was also slow. I'm confused - why can't this simple task be quick?
I spent a life solving this problem, so answer would be appreciated.
|
|
|
|
|
Here are some possible suggestions i've encountered with similar Stuff:
- Is the OnPaint method called very often? (stupid thing i once did number 1  )
- Try using a label instead of a panel. A label is double-buffered (if i remember right) and a panle isn't. Better yet, make your own custom control with the styles DoubleBuffer, AllPaintingInWmPaint and UserPaint on.
- Make sure the source rectangle and the destination rectangle are the same width and height, so the GDI doesn't have to schale the screen image.
- It can also be a problem as a result of the big image (i think 7 MB is a lot). If it's possible, try drawing a bitmap with a pixel format of Format32bppPArgb, it is said to be the fastest.
I hope this helps!
"..Commit yourself to quality from day one..it's better to do nothing at all than to do something badly.."
-- Mark McCormick
|| Fold With Us! || Pensieve || VG.Net ||
|
|
|
|
|
In addition to what Marc said...
Instead of redrawing the image from the same 7M pixel image (huge!), have your control draw the thumbnail image to an internal buffer first, the same size as the preview window. Then when your preview has to draw, it can more quickly draw it from the internal buffer version, which should be pretty small.
I've had to do this a few times in some custom controls that I wrote. Instead of calculating the same image over and over again in order to redraw a control image that doesn't change much, I monitor the Resize events of my control and redraw the control image to an internal buffer. Then in the Paint event for the control, I draw the control image from the buffer.
For example, take a clock face. Most of what you see doesn't change, except for the hand positions. But, to get smoothly turning clock hands, you have to redraw the control about every 30-50 ms at least. Since calculating the positions of the various bits of a clock face can be time consuming, a clock face image should be calculated and drawn once to a cached image. When the control has to redraw itself (every 30ms), it merely has to draw the clock face from the cache, then calculate and draw only the hand positions. When the size of the control changes, you recalculate and redraw the cache image to match the size of the control.
RageInTheMachine9532
"...a pungent, ghastly, stinky piece of cheese!" -- The Roaming Gnome
|
|
|
|
|
In fact, I'm afraid that using buffer will enhance redrawing when image position doesn't change, but when user will need to shift image viewport by a "hand tool", the redrawing buffer with drawing it on control will take slightly more time than redrawing the control directly. However, the buffer idea is good, because there can be more OnPaint calls, not only on mouse moves...
Another idea was to use a three times larger buffer than the control. When user drags mouse over control, the cursor is locked inside the control bounds and OnPaint works just with the buffer. When user releases mouse button, the buffer will update. This can be a little frustrating, when user needs to scroll whole large image (more dragging), but this can be also fast.
I'll try these approaches, thank you very much for the new ideas  .
|
|
|
|
|
ltinka wrote:
Another idea was to use a three times larger buffer than the control. When user drags mouse over control, the cursor is locked inside the control bounds and OnPaint works just with the buffer. When user releases mouse button, the buffer will update. This can be a little frustrating, when user needs to scroll whole large image (more dragging), but this can be also fast.
Instead of tieing the user to a small section of the image per move, if they reach the edge of the bufffer cause an update then. A periodic delay would be much less bothersome for the user than having to repeatedly grab and let go to scroll a large distance.
|
|
|
|
|
I'm running an app that creates several hundred thousand files and directories. After it runs a few hours the system runs out of resources. I've tracked the problem own to a lack of PTEs.
When I googled this I saw responses on how to change registry settings to create more PTE space but this won't work for two reasons.
1) The number of files can be arbitrarily large.
2) This is a consumer product and I can't expect the consumer to tweak the registry.
Has anyone else seen this problem or have an idea what is going on?
|
|
|
|
|
This doesn't really make any sense. PTE are used by the Virtual Memory Manager. This shouldn't affect the number of files/folders you create on a volume.
Are you looking at a memory or handle leak perhaps?
You could increase this pool, but your app would have to modify the registry (preferrably during installation) and restart for the changes to take effect.
I'd test the crap out of this solution though. By increasing the PTE pool, you could be merely postponing the inevitable because of leaky code.
RageInTheMachine9532
"...a pungent, ghastly, stinky piece of cheese!" -- The Roaming Gnome
|
|
|
|
|
I originally discovered what was causing my problem by doing a !vm in the kernel debugger. There was no problem with virtual memory except for an upper-case message: "YOU ARE OUT OF PTES".
I know that my handles are ok. Both the task manager and the openfiles command verify it.
In my original message I said I was creating hundreds of thousands of files. I was mistaken. I am actually creating millions of files, each about 256 kbytes. I structured the directories so that there are no more than 10,000 files per directory.
I also forgot to say that I am using interop from C# to do normal CreateFiles and WriteFiles from C. Also, I am doing the reads from a volume snapshot in xp sp2.
I did a test of running my code with the actual file-writing disabled and had no problem. All I can figure is that this is a bug in windows when you write too many files. I guess I will have to change my design to reduce the actual number of files written.
|
|
|
|
|
Well, the limit of the number of files in a single NTFS volume, depending on disk size of course, is 4.294 Billion. On a FAT32 volume, it's 4.17 Million. It's not a bug in Windows, but a problem with I/O capacity and the fact the you're probably one of the few people on earth to generate this many files on a volume. God help you if CHKDSK ever has to run. I hope you can take a week off...
It is recommended that, on volumes that contain an excessively large number of files, and Millions  qualifies, that you go through this[^] document on MSDN. Disable 8.3 name generation and also turn off Last Access timestamp updates. You'll probably have to do more than that...
RageInTheMachine9532
"...a pungent, ghastly, stinky piece of cheese!" -- The Roaming Gnome
|
|
|
|
|
I wanted to know if given an ip range, i.e 192.168.1.0-192.168.0.255, how do I find out which IPs are actual machines connected to the LAN without having to ping each IP?
Even if I ping each IP, I wouldn't know if a timed out IP is an actual machine connected the network but offline or if the IP doesn't even exist on the LAN.
If anyone has used the program Look@Lan, they would know that when you input an IP range it automatically figures out which IPs are valid and discards the ones that don't exist on the LAN. How do I accomplish this using C#?
I'm using Visual Studio .NET 2005 with framework 2.0.
I would appreciate any help or advice.
Thanks.
|
|
|
|
|
There's really two questions you seem to be confusing here. The first is "which IPs are valid within a range", and "which IPs are used by computers".
The first question can be solved mathematically, but it won't tell you anything about your actual network. The second question can only be solved by sending a ping or ARP packet out for each address, and checking for a response. Whether a machine is disconnected or simply nonexistant really is the same thing as far as network is concerned, so there's no way (or reason) to differentiate between the two.
If you want to know which addresses within a range of IPs have computers associated, ping each one. It's tedious, but it's the only way. You'll obvously only want to scan a small range of addresses, or you'll be at it for hours.
|
|
|
|
|
Thx for the reply tylerl.
I was hoping there was some other way to find out which IPs are used by computers besides pinging each one.
I don't know if you have used Look@Lan, but somehow it figures out which IPs are used by computers and which IPs just don't exist. I've tried looking around the net for the program's source code but even though its free, the source code is unavailable. I've also looked at quite a few ping applications in c# but they still can't differentiate between an IP that is being used by a computer or one that doesn't exist on the network. They just send a packet and check for a reply.
If you can think of anything else, please don't hesitate to reply back.
Thanks again.
|
|
|
|
|
Pinging is no longer much help. Unless you turn on the File Sharing exception for Windows Firewall in XP SP2, or otherwise open port 445, Windows Firewall will reject ICMP Echo Request packets sent by ping, by default. You can enable them if required. You'll find that most other software firewalls will do the same.
If you need to determine whether you can connect to a network service, just try to connect to it. If you're trying to do a site survey you're probably out of luck.
Stability. What an interesting concept. -- Chris Maunder
|
|
|
|
|
What is the best way to create one in C#?
I guess I need to create a tuple class or is there something like SortedList for tuples?
Later, JoeSox
"Football is a game of cliches, and I believe in every one of them." -Vincent Lombardi
CPMCv1.0 ↔ humanaiproject.org ↔ Last.fm
|
|
|
|
|
what do you mean by tuple, can you give an example? If you mean a variable that holds two values, you can easily use a struct or class
regards
modified 12-Sep-18 21:01pm.
|
|
|
|
|
Greeeg wrote: what do you mean by tuple, can you give an example?
Well, I am experimenting with this right now
<br />
class Tuple<br />
{<br />
private string relationship;<br />
private string phrase1;<br />
private string phrase2;<br />
<br />
public string Relationship<br />
{<br />
get { return relationship; }<br />
set { relationship = value; }<br />
}<br />
public string Phrase1<br />
{<br />
get { return phrase1; }<br />
set { phrase1 = value; }<br />
}<br />
public int Phrase2<br />
{<br />
get { return phrase2; }<br />
set { phrase2 = value; }<br />
}<br />
}<br />
I might change it to objects instead of strings. I just need something that will be fast for searching and sorting it's data. SortedList works great but that is only for a (key,value) not (key, value, value).
Later, JoeSox
"Football is a game of cliches, and I believe in every one of them." -Vincent Lombardi
CPMCv1.0 ↔ humanaiproject.org ↔ Last.fm
|
|
|
|
|
That's okay. The easiest way is to implement IComparable in your Tuple class. The SortedList will then sort your keys according to the result of the CompareTo function of your Tuple class.
regards
modified 12-Sep-18 21:01pm.
|
|
|
|
|
Thanks. I'll play around with that idea.
Later, JoeSox
"Football is a game of cliches, and I believe in every one of them." -Vincent Lombardi
CPMCv1.0 ↔ humanaiproject.org ↔ Last.fm
|
|
|
|
|
|
Kevin McFarlane wrote: If you're using .NET 2.0 then a Tuple class (actually a Triple)is provided by the free PowerCollections library.
Outstanding. Thanks.
Later, JoeSox
"Football is a game of cliches, and I believe in every one of them." -Vincent Lombardi
CPMCv1.0 ↔ humanaiproject.org ↔ Last.fm
|
|
|
|
|
Actually, I just discovered that there are Pair and Triplet classes of Objects in .NET 1.1 but they're in System.Web.UI. Surprised they didn't update them to generic versions and put them in the generics namespace. However, I expect they'll roll in some of the PowerCollections stuff over time.
Kevin
|
|
|
|
|
|
Has anybody ever drawn the red squigly lines that show up say in Word or other apps under a word when it is misspelled?
I have come up with a way but it is not reliable to underline all words. My method gets messed up with tab chars and also hidden text. I suspect the same would be true for tables and inserted pictures.
Thanks in advance,
Eric
|
|
|
|
|
From my knowledge, it is not directly possible to use the rtb class to draw the Office-Style underlines.
However, you can have an effective selection method by saving the selected text and creating custom functions to get text from it - even with hidden text
|
|
|
|
|
I think you would probably have to directly draw the lines on the control, either using RichTextBox.CreateGraphis() or by creating a class that inherits from RichTextBox, and overriding RichtextBox.OnPaint
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
