/ [CNS] / [Comp Sci Dept] / [R J Botting] >> [CSci202] >> 11
[Index] [Schedule] [Syllabi] [Text] [Labs] [Projects] [Resources] [Search] [Contact] [Grading]
Notes: [01] [02] [03] [04] [05] [06] [07] [08] [09] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20]
Tue May 8 07:08:47 PDT 2007

CSci202 Computer Science II, Session 11 Files and Strings

main

10.html

Preparation

tss2004.cpp

Study this page and 11.6( binary files) and 11.8 (stringstreams) and 11.9 (strstreams) and write down the questions, doubts, and surprises that you have on a piece of paper.

Assigned Work Due

Input

Looking inside files on UNIX -- file wc cat od

Down load [ example ] and save as source and then try these commands

The file command will make a good guess at the kind of file you have:

 	file example

The wc command will count the lines, words, and characters in a file:

 	wc -l example

 	wc -w example

 	wc -c example

The cat command outputs text files to the terminal (among other things).

 	cat example

The od command will tell you precisely what is in a file -- character by character or byte by byte. It makes unprintable and "white space" characters explicit.

 	od -c example

 	od -x example

 	od -cx example

My Experience of Binary files

only

The book uses direct access into a normal line-oreinted text file. I personally avoid designing programs and software that does this. I prefer handling irregular lines by one line after another... It is much simpler. But here is an example -- a program to read any 10 characters from a file: [ 11peek.cpp ] (Down load and try it.... there may be bugs).

Unformatted data is efficient because our program can go directly to any random place in the file: direct access or in IBM-speak "DASD" (dazz-dee). We have seek to move to a character in a random access file and tell to find out where the file was last at. Call them like this:

 		fstreamvar.tellp();

 		fstreamvar.seekp(position);

 		fstreamvar.tellg();

 		fstreamvar.seekg(position);

Demonstration of seek and tell with formatted data [ 11seek.cpp ]

But to be able to use direct access your program must be able to calculate where the data is.

Start by designing the layout of the data in the unformatted file. Draw a simple picture. Tabulate the data and calculate where it is in a simple case. Work out formulas that calculate where the general data is and hence code for reading and writing it. Document any special constraints. For example the file in lab06 must always start with zero or more real records followed by zero or more blank records. Another example would be to require that the file is always sorted in a particular way. A third trick is to place data near a position calculated by a hash function, that way it is quick to find it again.

It is also wise to not mix formatted input/output with unformatted input output. As a rule use 'read' and 'write' to handle the data, Stick to the following:

 		fstreamvar.read(address_of_object, sizeof(object)).

 		fstreamvar.write(address_of_object, sizeof(object)).

passwd.dat

11read.cpp

11write.cpp

Another rule that preserves sanity: never, ever switch from reading to writing without first executing a seekp. Similarly when you have written some data and want to read from a direct access file, always take care to do a seekg before reading. However you can safely execute a series of reads (or a series of writes) and get each item from the file in turn.

A direct access file is a repository for pieces of RAM. We can store them, keep them and then read them back in again. But when read they are are not necessarily any where near their old address.

Never store pointers or pointer based data in an unformatted file. This means that you can not store any of the Standard Library classes (vectors, strings, deques, etc.) in objects that are written and read into unformatted files.

It is better to use a data base to handle persistant data than invent your own data format for a direct access file!

More Experience with Random Access Files

Reason: the updates to the Gnu compiler changed the internal coding of objects like the user password in the lab. I guess that it added typeid information at the start of the object. So, what my code though was a length was something else....

The solution was to reconstruct the "passwd" file from scratch using newly compiled programs thruout.

Now, I'm hoping that they do not upgrade the compiler before the next lab.

So, add another reason for avoiding direct access binary files. The format of the data is compiler dependent.

stringstream and strstsream -- real cool stuff

tss2004.cpp

Handy for converting an argument to the main program into a numeric value. Also useful for interpreting a buffer of direct access data, or placing numeric data into character format in a direct access file.