I am looking for the best TEXT compression algorithm available

[Gotty]

You cannot get better than 87.5% if the data is random and the two symbols have equal probability.

"the data is random" entails "the two symbols have equal probability". You don't need both requirements.
I used the wording "truly random" to try to emphasize that but "the two symbols have equal probability" is probably better for CompressMaster to comprehend it. Add "no useful patterns" to the requirements, and you have the practical definition of "randomness".

Anyway, we are on the same message and thanx for the analysis.

[Gotty]

"QQQQQQq" will be replaced by "dog "
"QQQQq" will be replaced by "car "
"QQq" will be replaced by "cat "
and so on...

Text models usually help text compression, but they are tuned on real texts so they rely on properties of real texts ...

@Piotr: Thanx.
@CompressMaster: Replacing binary symbols or sequences with words will not make the file a real text file.
This is a real text file (string): "A dog was chasing a cat in a warm Saturday morning." This has the properties of real text. Text compressors are tuned for this. (See Piotr's post.)
This is a file containing binary information: "cat cat cat cat dog cat cat dog car". It does not have the properties of real text. Text compressors will not like it very much.
Conclusion: you cannot convert your binary file to a real text file.

[Gotty]

CompressMaster, you were right at the very beginning.
RLE preprocessing. Maximum 4-bit lenghts. Symbols are: series of O's, ending with one o.
(Why don't you tell us, how you generated the file?)

C#:

using System.Collections.Generic;
using System.IO;

namespace Decoder
{
    class Program
    {
        static void Main(string[] args)
        {
            //reading
            byte[] bytes = File.ReadAllBytes("only-O.txt");
            List<int> lengths = new List<int>();
            //unary to lengths
            int length = 0;
            for (int i = 0; i < bytes.Length; i++) {
                if (bytes[i] == 'O') length++;
                else {lengths.Add(length-1);length = 0;}
            }
            //encoding to nibbles (4-bits)
            List<byte> encoded_data = new List<byte>(lengths.Count);
            for (int i = 0; i < lengths.Count; i+=2)
              encoded_data.Add((byte)((lengths[i]<<4) | (lengths[i+1])));
            //writing result
            File.WriteAllBytes("photo.jpg",encoded_data.ToArray());
        }
    }
}

Edit: updated code to reveal photo immediately.

[Gotty]

On, it's already done in the other thread.
I'm late.

[mpais]

@Gotty
See here (JPEG nibble-encoded in unary with q and Q instead of 0 and 1).

132.827 bytes with paq8px_v151 after the bit1 tool from Shelwien to undo the encoding.

[Gotty]

Yeah, I'm kind of late. Nevertheless I updated my code to reveal the photo immediately.
You are right, it's unary-encoded.

[Kennon Conrad]

CompressMaster, you were right at the very beginning.
RLE preprocessing. Maximum 4-bit lenghts. Symbols are: series of O's, ending with one o.
(Why don't you tell us, how you generated the file?)

C#:

using System.Collections.Generic;
using System.IO;

namespace Decoder
{
    class Program
    {
        static void Main(string[] args)
        {
            //reading
            byte[] bytes = File.ReadAllBytes("only-O.txt");
            List<int> lengths = new List<int>();
            //unary to lengths
            int length = 0;
            for (int i = 0; i < bytes.Length; i++) {
                if (bytes[i] == 'O') length++;
                else {lengths.Add(length-1);length = 0;}
            }
            //encoding to nibbles (4-bits)
            List<byte> encoded_data = new List<byte>(lengths.Count);
            for (int i = 0; i < lengths.Count; i+=2)
              encoded_data.Add((byte)((lengths[i]<<4) | (lengths[i+1])));
            //writing result
            File.WriteAllBytes("photo.jpg",encoded_data.ToArray());
        }
    }
}

Edit: updated code to reveal photo immediately.

Stupid question, but I have to ask. Is that the all the code that's needed for C# and good practice, etc.? If so, I really need to learn it. It looks so nice and clean compared to C.

[Gotty]

Yeah, that's all the code needed to grab only-O.txt and convert it to photo.jpg.

{start of ad}

When you need to develop/test a new idea, it's perfect. From idea to working code it's just a couple of minutes. There are a lot of useful pre-baked classes (the .net environment). They are really useful.
C# makes/helps/forces you to think object oriented.
C# and java are very close relatives. If you learn one, you are really close to learn the other.

The memory management is automatic. No pointers, no memory leaks.
You can skip a lot of error handling like index boundary checks - the system will do it for you automatically.

Now it comes with a free development environment with a (really good) debugger, code editor, performance profiler, code analyzer, whatnot.
My favorite IDE feature (beside the debugger) is that when you type, and reach for variables, classes, objects, object members, the editor immediately tells you your choices. This feature is really helpful for fast development. Also you don't even have to compile your code to find out if you have some (stupid) mistakes: the editor tells you on the fly as you type what problems it sees.

It is worthy to learn it.

{end of ad}