lowlevellibsandfws/apputils/engineering/stringpool/StringPool.dox
author Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
Tue, 27 Apr 2010 18:23:16 +0300
branchRCL_3
changeset 20 a2e897c5c62b
parent 0 e4d67989cc36
permissions -rw-r--r--
Revision: 201012 Kit: 201017

/** @mainpage

	@section Header
	<center>
	<Table>
	<tr><td>Owner:</td><td>Application Frameworks & Protocols</td></tr>
	<tr><td>Author:</td><td>Leon Clarke</td></tr>
	<tr><td>Last Revised:</td><td>20 March 2002.</td></tr>
	<tr><td>Revision:</td><td>3.0</td></tr>
	<tr><td>Status:</td><td>Released</td></tr>
	<tr><td>Reviewers:</td><td></td></tr>
	<tr><td>Approval:</td><td>(PM) Urmi Shah, (ESM) Akin Oyesola, (Architect) Andrew Baldwin</td></tr>
	</Table>
	</center>
	<hr>

	@section Revision History
	<Table>
	<tr><td>Date</td><td>Version</td><td>Status</td><td>Description</td></tr>
	<tr><td>21-05-2001</td><td>1.0</td><td>Issued</td><td>Issued</td></tr>
	<tr><td>05-07-2001</td><td>1.1</td><td>Added more detail, in particular about static tables.</td><td>Issued</td></tr>
	<tr><td>05-07-2001</td><td>1.2</td><td>Minor editorial changes following T3 review comments.</td><td>Issued</td></tr>
	<tr><td>23-07-2001</td><td>1.3</td><td>Added reference to example code.</td><td>Issued</td></tr>
	<tr><td>09-11-2001</td><td>2.0</td><td>Added mutiple table support and case-sensitive strings</td><td>Issued</td></tr>
	<tr><td>20-03-2002</td><td>3.0</td><td>Incorporated changes following defect fixing</td><td>Released</td></tr>
	</Table>


	@section Introduction

   The string pool is a way of storing strings that makes comparison
   almost instantaneous at the expense of string creation. It is
   particularly efficient at handling string constants that are known
   at compile time. It currently only supports 8 bit strings. The
   basic algorithm is to ensure that the pool only contains one string
   of any particular value, using reference counts to keep track of
   it. Subsequent strings with the same value will actualy refer to
   the same copy.

   To use the string pool, you need an RStringPool object. Strings
   from different string pools can't be compared, so there should only
   be 1 string pool open per thread (unless one is in a component that
   doesn't export its use of the string pool). Of course, you can have
   multiple RStringPool objects, as long as they are all handles to
   the same string pool.

   Within this object, there are 2 distinct pools of strings, one
   comprising strings that are 'case-sensitive' (represented using the
   RString class) and another representing strings that are 'case-insensitive' (RStringF). Case-sensitive strings behave in a
   straightforward way. Case-insensitive strings should be used in a
   situation where case doesn't matter at all; if you create a new
   RStringF that differs from an existing RStringF only in terms of
   case, when you read back the value of the 'new' string, it will
   have the same value as the 'existing' one. This can be very useful
   in situations where strings are considered case-insensitive, but
   there is a 'traditional case' that is normaly used. As long as the
   first value to be added is in the 'traditional case', all
   subsequent additions will be corrected to match this first entry.

   Case-insensitivity assumes a character set of us-ascii (i.e.it only
   considers A-Z to be equivalent to a-z)

   If a string has a more complex or different case-sensitivity
   requirement that this model doesn't match, then it may be necessary
   to compare strings outside the pool, which may make the string pool
   inapplicable.

   Corresponding to RString and RStringF are the classes RStringToken
   and RStringTokenF. These are smaller (4 bytes long rather than 8)
   but need to be turned into String or StringF classes before you can
   do anything useful with them. In particular, you can't directly
   find the contents of a string token. They should be used when space
   is absoluteley at a premium, for instance storing a lot of strings
   in an array, or similar applications.

   @section Static Tables

   A very important aspect of the string pool is the concept of static string
   table. This is best thought of as an array of common
   strings. Integers can be transformed into the strings corresponding
   to that index position in the array via the String and StringF
   functions. In addition, a string can be cast back to an integer,
   allowing a switch statement on strings (it returns -1 if the string
   isn't one from any static table of the pool). Another advantage to using the
   static table is that the function to create the string can't leave,
   and the string doesn't need to be closed (although closing it is
   harmless)

   Multiple case-insensitive and case-sensitive string tables are supported. 

   A static table is written as a .st file, and is processed by
   stringtable.pl early in the build process to generate the actual
   cpp and header files at compile time. The format of the .st file is
   basicaly as follows. The first noncomment line consists of
   fstringtable <TableName> for a case-insensitive table and stringtable <TableName>
   for a case-sensitive table. Each subsequent non-comment line consists
   of the name of an enum followed by the value of the string. Comment
   lines start with # and can be added anywhere. You can also have
   lines starting in !, which will be output into the header file if
   you want additional comments in the header. e.g.

   @code
   # Example String Table
   fstringtable ExampleStringTable
   !// Some types of fruit
   # This comment won't appear in the .h file, but the one above is.
   EApple apple
   EOrange orange
   EBanana banana
   # Some animals
   ECat cat
   EDog dog
   @endcode

   The string table name must be a valid C++ class name, and the
   generated code includes a class that contains within it an enum
   corresponding to the elements in the array, so entries in the above
   example could be referred to as MyStringTable::ECat and so on.

   The stringpool.pl script takes a .st file and creates .cpp and .h
   files of the same name in the same directory. The easiest way to
   use this is to export the .st file to /epoc32/generated/, create an
   extension makefile that runs the script during the makefile phase
   and then copies the generated .h file to epoc32, and then compile
   the .cpp file as normal from an mmp files. Look at the example code
   in //EPOC/main/generic/bafl/docs/stringtableexample/... for a
   simple example of how to do this.

   To be notified when the String Pool is closing you can derive from the mix-in class    MStringPoolCloseCallBack and implement the StringPoolClosing() function. The OpenL(const    TStringTable& aTable, MStringPoolCloseCallBack& aCallBack) overridden function must be    used to register the callback with the String Pool. Where aCallback is the pointer to the    callback. The StringPoolClosing() function will then get called when the String Pool is    closing. 



 **/