GZipStream and DeflateStream Different Compressed Sizes

[sedwick]

I'm using .NET 2.0.50727 in VS 2005 Pro, ENU Service Pack 1 (KB926601). I've been experimenting with the System.IO.Compr ession classes GZipStream and DeflateStream, and I found it interesting to note that they'll create different-sized compressed files depending on the procedure you use to take in the source file's data. With one, almost every file I feed them for compression ends up significantly bigger. That code is below, almost verbatim per MS's MCTS Exam 70-536 self-paced training kit. Below it are some sample size results per file type.

        static void CompressFile(string toCompress, string toBeCompressed)
        {
            FileStream srcFile = File.OpenRead(toCompress);
            FileStream destFile = File.Create(toBeCompressed);
            GZipStream compStream = new GZipStream(destFile, CompressionMode.Compress);
            int lilByte = srcFile.ReadByte();
            while (lilByte != -1)
            {
                compStream.WriteByte((byte)lilByte);
                lilByte = srcFile.ReadByte();
            }
 
            compStream.Close();
            srcFile.Close();
            destFile.Close();
 
            //Sample Results, GZipStream/DeflateStream:
            //Text document: 1.98K --> 1.79K/1.77K
            //Word document: 140K  --> 227K
            //Bitmap:        123K  --> 168K
            //PDF document:  248K  --> 347K/346K
        }
 

This code, however, performed much better:

        static void CompressBetter(string toCompress, string toBeCompressed)
        {
            byte[] fileBytes = File.ReadAllBytes(toCompress);
            FileStream destFile = File.Create(toBeCompressed);
            GZipStream compStream = new GZipStream(destFile, CompressionMode.Compress);
            compStream.Write(fileBytes, 0, fileBytes.Length);
 
            compStream.Close();
            destFile.Close();
 
            //Sample Results, both:
            //Text document: 1.98K --> 1.28K
            //Word document: 140K  --> 21.6K
            //Bitmap:        123K  --> 20.6K
            //PDF document:  248K  --> 301K (Well, I guess they can't all work)
        }
 

Perhaps these classes like having more to work with up front? Anyone else have any opinions/observations?

[Plater]

Open the zip file in some program (like winzip or pkzip) and see what it says for compression values?

This like text files and bitmaps compress very well, PDF and other already-compressed data type don't compress very well.