Скачать Turbo BTree 1.6 - BTrees Implementation for TP4-5.5

27.07.1989
Скачать файл (206,35 Кб)

( Turbo BTree version 1.6 )

   Yet another implementation of BTrees for Turbo Pascal 4.0/5.0/5.5

               Copyright (c)  1988,1989    Dean H. Farwell II


Anyone using this software to develop a commercial application must
register and state their intentions to use it in a commercial
application.  I will not charge royalties (except for the registration
fee).  I just want to know what people are doing with it.
Also, if someone is distributing it as part of a shareware product,
the same restrictions apply.

Most of all I want any feedback that you have the time and desire to
submit. Suggestions, complaints, etc. are welcome.  I especially want
feedback on errors.  I do not know of any bugs (or I'd fix them) but
I probably missed an item or two.  Send me notice immediately so that
I can distribute a fix if required.  Since you have the source code,
do your best to isolate the problem.  I'll try to take it from there.

To reach me use CompuServe ID - 73240,3335 or drop me a line at:

                            P.O. Box 4731
                            San Bernardino, CA 92409

                            (zip code is essential!!!!!)

My phone number is:

                                714-887-9847


                             Product Description

TBTREE16 is version 1.6 of a shareware btree and database product for
use with Turbo Pascal 4.0, Turbo Pascal 5.0 and Turbo Pascal 5.5.
It is intended to give a user the capabilities to develop pascal
applications for the creation and manipulation of databases.
This product is extremely powerful and FAST!!!
However, it is not the easiest thing to get a grasp on.
If you understand the concept of btrees or at least indexing then
this will not be complicated to use.  I can just about guarantee that
it is the best product of its type for the money and maybe the best
period.  If you have any feedback on how I can take the mystery out
of using this please let me know.  I also solicit any feedback which
will enhance the quality of the product.

This product has been greatly enhanced since version 1.0 was released.
As it presently stands, it has the following features:

1. The ability to create data files and indexes for those files.

2. Any number of indexes can be used for a single data file.

3. Any number of data files can be used.

4. All data files and index files use a virtual memory scheme in which
   the most recently used pages are kept in memory using the heap.
   This is different than many other implementations which use a
   buffer or stack only for index pages.  The buffer is shared by
   all files.  In this way separate space does not need to be set
   aside for different files.  Space utilization is maximized and
   performance is enhanced.

5. All management of pages in memory is automatically controlled
   although the user can exercise limited control if desired.

6. Routines are provided to dump the page buffer to the printer to
   facilitate viewing pages if a user is curious or masochistic.

7. The size of the buffer is determined by the user and can be changed
   dynamically as the program runs.  For example a 128K buffer could
   be used initially.  If a user had a large heap requirement the
   buffer size could be reduced.  The buffer size could later be
   increased if desired.  This can all be done while the application
   is running.

8. Problems related to keeping track of numbers of open files are
   alleviated.
   The user initially specifies the maximum number of files allowed
   to be open at once and never has to address it again.
   Routines are provided to manage the opening and closing of files
   dynamically.  This increases performance because files do not have
   to be continually opened and closed unnecessarily.

9. All data types supported by Turbo Pascal are explicitly supported.
   In other words, values do not have to be converted to strings.
   This is a great improvement over previous systems
   (MY UNBIASED OPINION). For example an index can be set up for
   type Byte, another for type ShortInt, another of type String
   and even for type Real (but wait you say .. Reals in an index?).
   See next note!

10. Indexes can be searched for equality, inequality, greater than,
    greater than or equal to, less than, less than or equal to,
    and existence.  In other words you can ask for a list of all
    records for which some field of type real is LE (less than or
    equal to) XXX where XXX is a var of type real and XXX := 3.141592
    etc.  An index can also searched on a range of values.
    For instance an index of type byte can be checked for a
    range >= 10 and < 20 etc.  Indexes containing strings can be
    searched for partial string matches.  Partial string matches
    can be done three ways: find a string which is at the beginning
    of a string, anywhere in the string, or at the end of a string.

11.  As physical records are deleted space is reused automatically
     on future inserts.  Files should never have to be reorganized
     to recover space (unless a large number of deletes were
     performed without any inserts and even in this case there is
     not really a good reason to reorganize unless your disk is
     filling up).

12.  Included with the package is a very crude but effective source
     code printing routine which can be used to dump the code in a
     more usable format.  Specifically, it handles page breaks and
     prints a header of useful information.  It will also only print
     out the interface section if you instruct it to do so.

13.  The user does not need to perform any calculations or make any
     estimates prior to data file or index creation.

14.  Code was designed using an Object Oriented Design approach to
     a large extent (Turbo 4.0 and Turbo 5.0 lend themselves to this
     approach much more than Turbo Pascal 3.0 did).  This is one
     reason that there are so many units.  Much of the code is
     reusable for other applications.

15.  Sorts can be accomplished using any number of fields in any order
     of precedence as determined by you.  A Quicksort algorithm is
     used to give high performance.  Sort fields can be any type
     supported (Byte, Real, Integer, String etc.).

16.  Added the following set operations: Union, Intersection and
     Difference. These allow easy queries on multiple fields.
     For example, retrieving all personnel who have a last name
     starting with 'N' and who's job is 'MANAGER', etc. is greatly
     simplified.  This enhancement is a definite plus and not
     available in many other products.

17.  Error handling added for trapping I/O errors

18.  Capability to handle files with fixed size records and files
     with variable sized records