[SOLVED] Making a new folder, reading an entire file in one block read

[fogus]

Hello,

I am working on a new database indexing structure (to index a heap file). I have a need to create folders in which to store various index files.

Secondly, I need to be able to read my files (files, NOT folders) in as few hard disk head movements as possible. I need to pull them into memory and then execute a linear search on them, but I do not want to read them line by line if this would cause n reads for a file containing n lines.

Thanks,

[Webuser]

Use

new File("");

to solve that problem

[toadaly]

Use a FileInputStream as the source object of a Buffered InputStream.

BufferedInputStream in = new BufferedInputStream(new FileInputStream(...yoiur file name...));

[fogus]

Use
to solve that problem

Thanks.

If that is how I make a new folder, then how do I make a new file?

[fogus]

Use a FileInputStream as the source object of a Buffered InputStream.

BufferedInputStream in = new BufferedInputStream(new FileInputStream(...yoiur file name...));

Sounds good. So that grabs the whole file in one go? And then I can access it one line at a time? Nice.

[Webuser]

try it...

File f=new File("folder_1");
  f.mkdir();

[fogus]

OK, I will, for sure.

Another question: does anyone know how to make a "folder" structure within a single file ("single file" as in one .dat file in windows, for example) that can be accessed one "file" at a time? Is there a module to do something like that?

Cheers,

[toadaly]

You can create zip and jar files in Java. These can have an embedded directory structure. Is that what you mean?

[Webuser]

You can create zip and jar files in Java. These can have an embedded directory structure. Is that what you mean?

yeah ) It is quite good way as Jar file creation. Pack your project there. And there ia a code way to get data out the jar file by Java )

[fogus]

Well, what I am trying to do is create a new indexing method for a database (where records are stored in an unordered fashion in a heap file). My new structure is a mix between a hash table and a tree. It is extremely simple, easy to implement, and (I believe) efficient.

Perhaps all I need is more control over how my files are read and written. I need some kind of way to write to a file without having to shift all the contents around. My index file has to persist on the hard drive through program restarts.

Imagine this: you have a B+ tree index and you want to write it to the hard drive. How do you do it? I becha that if you can answer that question with sufficient detail, I will be able to figure out how to make my new structure persist.

Cheers,

[Webuser]

you can put all resources right into your jar file in a folder structure you need and then sinply get out the recources you need to. But there may cause a problem with jar in the case you need to input the data (as db may should) :(
What client you want to use. Is it an applet or an dt application?

[toadaly]

Perhaps all I need is more control over how my files are read and written. I need some kind of way to write to a file without having to shift all the contents around.

Serialize in memory, and then write out to disk as a byte array. Don't use the file system as RAM, and you should be ok.

[fogus]

"Serialize in memory"

That sounds cool, but I don't know how to do it. (sorry, I'm a beginner)

[toadaly]

Rather than writing directly to a file, write to a byte array. Then write the byte array to the file. If you explain a little more about how you're doing it with files, we can explain how to do it in memory.

[fogus]

I think my problem is not that I cannot build the objects that I need in memory; it is that I do not know how to make those objects persist across program restarts.

I need to have my index structure (think B+ tree) saved to a file so that it does not vaporise when I shutdown my program.

I have made a binary tree before, and that was all well and good, but every time I needed to use it for an index I would have to rebuild it (because it disappeared last time I shut the program down).

The other thing is, I can only assume an arbitrarily large hard drive, not RAM. I would much rather store my index structure on the hard disk (hence the similar structure to B+ trees, structures designed for a minimum of disk reads). Does anyone know how to make a B+ index in Java? I can't think of a way to make one persist. Where do I store the leaves? Do I need a lot of separate files or can I use one single large file?

[fogus]

Has anyone built a persistent B+ tree that is simple enough to explain the workings of to me?