We Keep Coding

Add data to a zip archive until it reaches a given size

I'm trying to zip some data, but also break up the data set into multiple archives so that no single zip file ends up being larger than some maximum.
Since my data is not sourced from the file system it seems a good idea to use a streaming approach. I thought I could simply write one atomic piece of data at a a time while keeping track of the stream position prior to writing each piece. Once I exceed the limit, I truncate the stream to the position before writing the piece that can't fit, and move on to create the next archive.
I've attempted with the classes in System.IO.Compression - create an archive, create an entry, use ZipArchiveEntry.Open to get a stream, and write to that stream. The problem is it seems impossible to get from this how large the archive is at any point.
I can read the stream's position, but this is tracking uncompressed bytes. Truncating the stream works fine too, so I have this working as intended now with the important exception that the limit applies to how much uncompressed data there is per archive rather than how large the compressed archive becomes.
The data is part compressible text and various blobs (attachments from end users) that will sometimes be very compressible and sometimes not at all.
My questions:
1) Is there something about the deflate algorithm that inherently conflicts with my approach? I know it is a block-based compression scheme and I imagine the algorithm may not decide how to encode the compressed data until the entire archive has been specified.
2) If the answer to (1) above is "yes", what is a good strategy that doesn't introduce far too much overhead?
One idea I have is to assume that compressed data won't be larger than uncompressed data. I can then write to the stream until uncompressed data exceeds the threshold, and then save the archive, calculate the difference between the threshold and the current size, and repeat until full.
In case that wasn't clear, say the limit is 1MB. I write 1 MB of uncompressed data and save the archive. I then see that the resulting archive is 0.3MB. I open the archive (and its only entry) again and start over with a new limit of 0.7 MB, since I know I am able to add at least that much uncompressed data to it without overshooting. I imagine this approach is relatively simple to implement, and will test it, but am interested to hear if anyone has better ideas.

You can find out a lower bound on how big the compressed data is by looking at the Length or Position of the underlying FileStream. You can then decide to stop adding entries. The ZIP stream classes tend to buffer not too much. Probably on the order of 64KB.
Truncating the archive at a certain point should be possible. Try flushing the ZIP stream before measuring the Position of the base stream. This is always possible in theory but the actual library you are using might not support it. Test it or look at the source.

Combining FileStream and MemoryStream to avoid disk accesses/paging while receiving gigabytes of data?

I'm receiving a file as a stream of byte[] data packets (total size isn't known in advance) that I need to store somewhere before processing it immediately after it's been received (I can't do the processing on the fly). Total received file size can vary from as small as 10 KB to over 4 GB.
One option for storing the received data is to use a MemoryStream, i.e. a sequence of MemoryStream.Write(bufferReceived, 0, count) calls to store the received packets. This is very simple, but obviously will result in out of memory exception for large files.
An alternative option is to use a FileStream, i.e. FileStream.Write(bufferReceived, 0, count). This way, no out of memory exceptions will occur, but what I'm unsure about is bad performance due to disk writes (which I don't want to occur as long as plenty of memory is still available) - I'd like to avoid disk access as much as possible, but I don't know of a way to control this.
I did some testing and most of the time, there seems to be little performance difference between say 10 000 consecutive calls of MemoryStream.Write() vs FileStream.Write(), but a lot seems to depend on buffer size and the total amount of data in question (i.e the number of writes). Obviously, MemoryStream size reallocation is also a factor.
Does it make sense to use a combination of MemoryStream and FileStream, i.e. write to memory stream by default, but once the total amount of data received is over e.g. 500 MB, write it to FileStream; then, read in chunks from both streams for processing the received data (first process 500 MB from the MemoryStream, dispose it, then read from FileStream)?
Another solution is to use a custom memory stream implementation that doesn't require continuous address space for internal array allocation (i.e. a linked list of memory streams); this way, at least on 64-bit environments, out of memory exceptions should no longer be an issue. Con: extra work, more room for mistakes.
So how do FileStream vs MemoryStream read/writes behave in terms of disk access and memory caching, i.e. data size/performance balance. I would expect that as long as enough RAM is available, FileStream would internally read/write from memory (cache) anyway, and virtual memory would take care of the rest. But I don't know how often FileStream will explicitly access a disk when being written to.
Any help would be appreciated.

No, trying to optimize this doesn't make any sense. Windows itself already caches file writes, they are buffered by the file system cache. So your test is about accurate, both MemoryStream.Write() and FileStream.Write() actually write to RAM and have no significant perf differences. The file system driver lazily writes it to disk in the background.
The RAM used for the file system cache is what's left over after processes claimed their RAM needs. By using a MemoryStream, you reduce the effectiveness of the file system cache. Or in other words, you trade one for the other without benefit. You're in fact off worse, you use double the amount of RAM.
Don't help, this is already heavily optimized inside the operating system.

Since recent versions of Windows enable write caching by default, I'd say you could simply use FileStream and let Windows manage when or if anything actually is written to the physical hard drive.
If these files don't stick around after you've received them, you should probably write the files to a temp directory and delete them when you're done with them.

Use a FileStream constructor that allows you to define the buffer size. For example:
using (outputFile = new FileStream("filename",
FileMode.Create, FileAccess.Write, FileShare.None, 65536))
{
}
The default buffer size is 4K. Using a 64K buffer reduces the number of calls to the file system. A larger buffer will reduce the number of writes, but each write starts to take longer. Emperical data (many years of working with this stuff) indicates that 64K is a very good choice.
As somebody else pointed out, the file system will likely do further caching, and do the actual disk write in the background. It's highly unlikely that you'll receive data faster than you can write it to a FileStream.

Getting the file size of a file in C#

I want to write a file back to the client using a custom handler in ASP.Net and wondering what it the best way to do this with the least processing time. At the moment I have 2 different versions of the code that do the same thing, but because the handler will be used a lot, I'm wondering what the most efficient way is.
Load the complete file to a byte array and use BinaryWrite to write the file:
string filePath = context.Server.MapPath(context.Request.Url.LocalPath);
Byte[] swfFile = File.ReadAllBytes(filePath);
context.Response.AppendHeader("content-length", Utils.MakeString(swfFile.Length));
context.Response.ContentType = Utils.GetMimeType(Path.GetExtension(filePath));
context.Response.BinaryWrite(swfFile);
Using a FileInfo object to determine the file length and TransmitFile to write the file:
string filePath = context.Server.MapPath(context.Request.Url.LocalPath);
FileInfo fileInfo = new FileInfo(filePath);
context.Response.AppendHeader("content-length", Utils.MakeString(fileInfo.Length));
context.Response.ContentType = Utils.GetMimeType(Path.GetExtension(filePath));
context.Response.TransmitFile(filePath);
I would suspect the TransmitFile method is the most efficient because it writes without buffering the file. What about the FileInfo object? How does it calculate the file size? And is a FileInfo object the best way to do this or is there an better way?

FileInfo asks the filesystem for the file size information (it does not need to read all of the contents to do so of course). This is usually considered an "expensive" operation (in comparison to manipulating stuff in memory and calling methods) because it hits the disk.
However, while it's true that cutting down on disk accesses is a good thing, when you are preparing to read the full contents of a file anyway this won't make a difference in the grand scheme of things. So the performance of FileInfo itself is what you should be focusing on.
The #1 performance issue here is that the first approach keeps the whole file in memory for as long as the client takes to download it -- this can be a huge problem, as (depending on the size of the files and the throughput of the client connection) it has the potential of massively increasing the memory usage of your application. And if this increased memory usage leads to swapping (i.e. hitting the disk) performance will instantly tank.
So what you should do is use TransmitFile -- not because it's faster when going to the disk (it may or may not be), but because it uses less memory.

What is the best way to work with files in memory in C#?

I am building an ASP.NET web application that creates PowerPoint presentations on the fly. I have the basics working but it creates actual physical files on the hard disk. That doesn't seem like a good idea for a large multi-user web application. It seems like it would be better if the application created the presentations in memory and then streamed them back to the user. Instead of manipulating files should I be working with the MemoryStream class? I am not exactly sure I understand the difference between working with Files and working with Streams. Are they sort of interchangeable? Can anyone point me to a good resource for doing file type operations in memory instead of on disk? I hope I have described this well enough.
Corey

You are trying to make decision that you think impacts performance of your application based on "doesn't seem like a good idea" measurement, which is barely scientific. It would be better to implement both and compare, but first you should list your concerns about either implementations.
Here are some ideas to start:
there are really not much difference between temporary files and in-memory streams. Both would have content in physical memory if they are small enough, both will hit the disk if there is memory pressure. Consider using temporary Delete on close files for your files if cleaning files up is the main concern.
OS already doing very good job for managing large files with caching, one would need to make sure pure in-memory solution at least matches it.
MemoryStream is not the best implementation for reasonably sized streams due its "all data is in single byte array" contract (see my answer at https://stackoverflow.com/a/10424137/477420).
Managing multiple large in-memory streams (i.e. for multiple users) is fun for x86 platform, less of a concern for x64 ones.
Some API simply don't provide a way working with Stream-based classes and require physical file.

Files and streams are similar, yes. Both essentially stream a byte array...one from memory, one from the hard drive. If the API you are using allows you to generate a stream, then you can easily do that and serve it out to the user using the Response object.
The following code will take a PowerPoint memory object (you'll need to modify it for your own API, but you can get the general idea), save it to a MemoryStream, then set the proper headers and write the stream to the Response (which will then let the user save the file to their local computer):
SaveFormat format = SaveFormat.PowerPoint2007;
Slideshow show = PowerPointWriter.Generate(report, format);
MemoryStream ms = new MemoryStream();
show.Save(ms, format);
Response.Clear();
Response.Buffer = true;
Response.ContentType = "application/vnd.ms-powerpoint";
Response.AddHeader("Content-Disposition", "attachment; filename=\"Slideshow.ppt\"");
Response.BinaryWrite(ms.ToArray());
Response.End();

Yes, I would recommend the MemoryStream. Typically any time you access a file, you are doing so with a stream. There are many kinds of streams (e.g. network streams, file streams, and memory streams) and they all implement the same basic interface. If you are already creating the file in a file stream, instead of something like a string or byte array, then it should require very little coding changes to switch to a MemoryStream.
Basically, a steam is simply a way of working with large amounts of data where you don't have to, or can't, load all the data at into memory at once. So, rather than reading or writing the entire set of data into a giant array or something, you open a stream which gives you the equivalent of a cursor. You can move your current position to any spot in the stream and read or write to it from that point.

difference between memory stream and filestream

During the serialization we can use either memory stream or file stream.
What is the basic difference between these two? What does memory stream mean?
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Runtime.Serialization.Formatters.Binary;
namespace Serilization
{
class Program
{
static void Main(string[] args)
{
MemoryStream aStream = new MemoryStream();
BinaryFormatter aBinaryFormat = new BinaryFormatter();
aBinaryFormat.Serialize(aStream, person);
aStream.Close();
}
}
}

Stream is a representation of bytes. Both these classes derive from the Stream class which is abstract by definition.
As the name suggests, a FileStream reads and writes to a file whereas a MemoryStream reads and writes to the memory. So it relates to where the stream is stored.
Now it depends how you plan to use both of these. For eg: Let us assume you want to read binary data from the database, you would go in for a MemoryStream. However if you want to read a file on your system, you would go in for a FileStream.
One quick advantage of a MemoryStream is that there is not need to create temporary buffers and files in an application.

The other answers here are great, but I thought one that takes a really high level look at what purpose steams serve might be useful. There's a bit of simplification going on in the explanation below, but hopefully this gets the idea across:
What is a stream?
A stream is effectively the flow of data between two places, it's the pipe rather than the contents of that pipe.
A bad analogy to start
Imagine a water desalination plant (something that takes seawater, removes the salt and outputs clean drinking water to the water network):
The desalination plant can't remove the salt from all of the sea at one time (and nor would we want it to… where would the saltwater fish live?), so instead we have:
A SeaStream that sucks a set amount of water at a time into the plant.
That SeaStream is connected to the DesalinationStream to remove the salt
And the output of the DesalinationStream is connected to the DrinkingWaterNetworkStream to output the now saltless water to the drinking water supply.
OK, so what's that got to do with computers?
Moving big files all at once can be problematic
Frequently in computing we want to move data between two locations, e.g. from an external hard drive to a binary field in a database (to use the example given in another answer). We can do that by copying all of the data from the file from location A into the computer's memory and from there to to Location B, but if the file is large or the source or destination are potentially unreliable then moving the whole file at once may be either unfeasible or unwise.
For example, say we want to move a large file on a USB stick to a field in a database. We could use a 'System.IO.File' object to retrieve that whole file into the computer's memory and then use a database connection to pass that file onto the database.
But, that's potentially problematic, what if the file is larger than the computer's available RAM? Now the file will potentially be cached to the hard drive, which is slow, and it might even slow the computer down too.
Likewise, what if the data source is unreliable, e.g. copying a file from a network drive with a slow and flaky WiFi connection? Trying to copy a large file in one go can be infuriating because you get half the file and then the connection drops out and you have to start all over again, only for it to potentially fail again.
It can be better to split the file and move it a piece at a time
So, rather than getting whole file at once, it would be better to retrieve the file a piece at a time and pass each piece on to the destination one at a time. This is what a Stream does and that's where the two different types of stream you mentioned come in:
We can use a FileStream to retrieve data from a file a piece at a time
and the database API may make available a MemoryStream endpoint we can write to a piece at a time.
We connect those two 'pipes' together to flow the file pieces from file to database.
Even if the file wasn't too big to be held in RAM, without streams we were still doing a number or read/write operations that we didn't need to. The stages we we're carrying out were:
Retrieving the data from the disk (slow)
Writing to a File object in the computer's memory (a bit faster)
Reading from that File object in the computer's memory (faster again)
Writing to the database (probably slow as there's probably a spinning disk hard-drive at the end of that pipe)
Streams allow us to conceptually do away with the middle two stages, instead of dragging the whole file into computer memory at once, we take the output of the operation to retrieve the data and pipe that straight to the operation to pass the data onto the database.
Other benefits of streams
Separating the retrieval of the data from the writing of the data like this also allows us to perform actions between retrieving the data and passing it on. For example, we could add an encryption stage, or we could write the incoming data to more than one type of output stream (e.g. to a FileStream and a NetworkStream).
Streams also allow us to write code where we can resume the operation should the transfer fail part way through. By keeping track of the number of pieces we've moved, if the transfer fails (e.g. if the network connection drops out) we can restart the Stream from the point at which we received the last piece (this is the offset in the BeginRead method).

In simplest form, a MemoryStream writes data to memory, while a FileStream writes data to a file.
Typically, I use a MemoryStream if I need a stream, but I don't want anything to hit the disk, and I use a FileStream when writing a file to disk.

While a file stream reads from a file, a memory stream can be used to read data mapped in the computer's internal memory (RAM). You are basically reading/writing streams of bytes from memory.

Having a bitter experience on the subject, here's what I've found out. if performance is required, you should copy the contents of a filestream to a memorystream. I had to process the contents of 144 files, 528kbytes each and present the outcome to the user. It took 250 seconds aprox. (!!!!). When I just copied the contents of each filestream to a memorystream, (CopyTo method) without changing anything at all, the time dropped to approximately 32 seconds. Note that each time you copy one stream to another, the stream is appended at the end of the destination stream, so you may need to 'rewind' it prior to copying to it. Hope it helps.

In regards to stream itself, in general, it means that when you put a content into the stream (memory), it will not put all the content of whatever data source (file, db...) you are working with, to the memory. As opposed to for example Arrays or Buffers, where you feed everything to the memory. In stream, you get a chunk of eg. file to the memory. When you reach the end of chunk, stream gets the next chunk from file to the memory. It all happens in low-level background while you are just iterating the stream. That's why it's called stream.

A memory stream handles data via an in memory buffer. A filestream deals with files on disk.

Serializing objects in memory is hardly any useful, in my opinion. You need to serialize an object when you want to save it on disk. Typically, serialization is done from the object(which is in memory) to the disk while deserialization is done from the saved serialized object(on the disk) to the object(in memory).
So, most of the times, you want to serialize to disk, thus you use a Filestream for serialization.