The PLVtkn (PL/Vision ToKeN) package determines whether an identifier is a PL/SQL keyword. It does this by using a token table containing these keywords.
Every PL/SQL program is filled with identifiers. Identifiers are named PL/SQL language elements and include variable names, program names, and reserved words. Reserved words play a very different role in our programs than do the application-specific identifiers. I recommend strongly in Oracle PL/SQL Programming that you reflect these different roles in your program by using the UPPER-lower method: all reserved words are typed in UPPER case and all application-specific identifiers are typed in lower case. I even go so far, in PL/Vision, as to provide you with a package (PLVcase) which will automatically convert your programs to the UPPER-lower method.
Well, if PLVcase is going to uppercase only keywords, it has to know which identifiers in a PL/SQL program are the reserved words. This information is maintained in the PLV_token table, which has the following structure:
Name Null? Type ------------------------------- -------- ---- TOKEN NOT NULL VARCHAR2(100) TOKEN_TYPE VARCHAR2(10)
where the token column is the identifier and token_type indicates the type.
The different token types in the PLV_token table are stored in the PLV_token_type table, which has this structure:
Name Null? Type ------------------------------- -------- ---- TOKEN_TYPE VARCHAR2(10) NAME NOT NULL VARCHAR2(100)
The contents of the PLV_token_type are explained in the following table:
Name | Description | |
---|---|---|
B | BUILT-IN | Builtin functions and procedures of the PL/SQL language, including packaged builtins. |
D | DATATYPE | Different datatypes of the PL/SQL language, such as INTEGER and VARCHAR2. |
DD | DATA-DICTIONARY | Views and tables from the Oracle Server data dictionary, such as ALL_SOURCE and DUAL. |
E | EXCEPTION | Predefined system exceptions such as ZERO_DIVIDE. |
OF | ORACLE-FORMS | Reserved words from the Oracle Forms product set |
S | SYMBOL | Symbols like + and =. |
SQL | SQL | Elements of the SQL language. In many cases, SQL tokens are used in PL/SQL and also in Oracle Developer/2000. These are still listed as SQL tokens. |
X | SYNTAX | Syntax elements of the PL/SQL language, such as AND or LIKE. |
There is a row in PLV_token for each reserved word in PL/SQL. You can change the contents of this table if you want. You might, for example, want to add keywords for the Oracle Developer/2000 builtins or the Oracle Web Agent PL/SQL packages. You can even add your own application-specific identifiers to the table. As long as the token type you assign is not any of those listed above, PL/Vision will not misinterpret your entries.
There are currently 1,235 rows in the PLV_token table, broken down by token type as follows:
Token Type | Count |
---|---|
BUILT-IN | 198 |
DATATYPE | 22 |
DATA-DICTIONARY | 168 |
EXCEPTION | 15 |
ORACLE-FORMS | 623 |
SYMBOL | 32 |
SQL | 94 |
SYNTAX | 83 |
From the PL/SQL side of things, the PLVtkn package provides an interface to the PLV_token table. This package is used by PLVcase to determine the case of an individual token according to the UPPER-lower method.
As you will soon see, PLVtkn is not a particularly large or complicated package. Its purpose in life is to consolidate all of the logic having to do with individual PL/SQL tokens, particularly regarding keywords. By hiding the implementation details (the name and structure of the PLV_token table, the particular values used to denote a symbol or syntax element or builtin function), PLVtkn makes it easier for developers to apply this information in their own programs.
PLVtkn provides a set of functions you can use to determine a string's token type in PL/SQL. The headers for these functions are shown below:
FUNCTION is_keyword (token_in IN VARCHAR2, type_in IN VARCHAR2 := c_any) RETURN BOOLEAN; FUNCTION is_syntax (token_in IN VARCHAR2) RETURN BOOLEAN; FUNCTION is_builtin (token_in IN VARCHAR2) RETURN BOOLEAN; FUNCTION is_symbol (token_in IN VARCHAR2) RETURN BOOLEAN; FUNCTION is_datatype (token_in IN VARCHAR2) RETURN BOOLEAN; FUNCTION is_exception (token_in IN VARCHAR2) RETURN BOOLEAN;
All of the functions except for is_keyword take a single string argument and return TRUE if the string is that type of token. The following examples illustrate the way the PLVtkn functions interpret various strings:
SQL> exec p.l(PLVtkn.is_builtin('to_char')); TRUE SQL> exec p.l(PLVtkn.is_builtin('loop')); FALSE SQL> exec p.l(PLVtkn.is_syntax('loop')); TRUE SQL> exec p.l(PLVtkn.is_syntax('=')); FALSE SQL> exec p.l(PLVtkn.is_symbol('=')); TRUE
The is_keyword function is a more general-purpose function. It returns TRUE if the token is a keyword of the type specified by the second argument. The default value for this second parameter is PLVprs.c_any, which means that is_keyword will return TRUE if the specified token is any kind of keyword.
PLVcase uses the is_keyword to determine whether the token should be upper- or lowercase. When applying the UPPER-lower method, it doesn't matter if the token is a builtin function or a syntax element, such as the END statement. All such keywords must be uppercase. Here is the code from the PLVcase.token procedure which performs the actual conversion:
IF PLVtkn.is_keyword (v_token) THEN v_token := UPPER (v_token); ELSE v_token := LOWER (v_token); END IF;
To keep code volume in PLVtkn to an absolute minimum and eliminate redundancy, I implement all of the "specialized" is functions (is_builtin, is_syntax, etc.) with a call to is_keyword, as shown below:
FUNCTION is_symbol (token_in IN VARCHAR2) RETURN BOOLEAN IS BEGIN RETURN (is_keyword (token_in, c_symbol)); END;
You will use the get_keyword procedure to retrieve from the PLV_token table all information stored about a particular token. The header of this procedure is:
PROCEDURE get_keyword (token_in IN VARCHAR2, kw OUT kw_rectype);
You provide the token or string and get_keyword returns a PL/SQL record, which is a translated version of the row in the table. The translation generally involves converting string constants to Boolean values. For example, one of the record's fields is named is_keyword. The expression assigned to this Boolean field is:
kw.is_keyword := kw_rec.token_type IN (c_syntax, c_builtin, c_symbol, c_sql, c_datatype, c_datadict, c_exception);
where kw_rec is the cursor-based record into which the PLV_token row is fetched.
The anonymous block below shows how to use get_keyword. It accepts a string from the user of this script (plvtkn.tst), retrieves the information about that string (as a token), and displays some of the data.
DECLARE my_kw PLVtkn.kw_rectype; BEGIN PLVtkn.get_keyword ('&1', my_kw); p.l (my_kw.token_type); p.l (my_kw.is_keyword); END; /
The lines below show this script being executed for the THEN keyword.
SQL> @plvtkn.tst then X TRUE
Copyright (c) 2000 O'Reilly & Associates. All rights reserved.