Update changelog.

[hcoop/debian/courier-authlib.git] / README.authmysql.myownquery

            Developer Notes for courier-imap-myownquery.patch




                                                        document version: 1.03
                                                        author: Pawel Wilk
















0 What's that?

1 Modifications overview

2 Definitions

3 New data types
  3.1 struct var_data
  3.2 typedef size_t (*parsefunc)
  
4 New functions
  4.1 get_variable
  4.2 parse_core
  4.3 ParsePlugin_counter
  4.4 ParsePlugin_builder
  4.5 parse_string
  4.6 validate_password
  4.7 get_localpart
  4.8 get_domain
  4.9 parse_select_clause
  4.10 parse_chpass_clause
  
5 Ideas and TODO

6 Thanks




                    *-----------------------
                     0 What's that?
                    *-----------------------

Courier-imap-myownquery.patch allows administrator to set own SQL queries
used by authdaemon to authenticate user (including fetchig credentials) and to
change user's password. It allows to construct SELECT or UPDATE clause in the
configuration file (authmysqlrc or authpgsqlrc) by adding two new configuration
variables: MYSQL_SELECT_CLAUSE or PGSQL_SELECT_CLAUSE and MYSQL_CHPASS_CLAUSE 
or PGSQL_CHPASS_CLAUSE. It may be useful in the mail environments where there
is such a need to have different database structure and/or tables scheme than
expected by authmysql or authpgsql module.

It also implements a small parsing engine for substitution variables which
may appear in the clauses and are used to put informations like username
or domain into the right place of a query.

This patch was created using `diff -Nur` on courier-imap-1.3.12 source.





                    *-----------------------
                     1 Modifications overview
                    *-----------------------

Modified files: authmysqllib.c authmysqlrc authpgsqllib.c authpgsqlrc

Each modified set of instructions is marked by my e-mail address:
siefca@pld.org.pl (for MySQL files) or tom@minnesota.com (for PostgreSQL files)

Changes in the current source code are related to:

- sections where the queries are constructed
  (including memory allocation for the buffers)

        when MYSQL_SELECT_CLAUSE or MYSQL_CHPASS_CLAUSE or 
        PGSQL_SELECT_CLAUSE or PGSQL_CHPASS_CLAUSE  is
        used then the query goes through the parsing functions
        passing over current memory allocation and query construction
        subroutines
          
- section where the configuration file is read

        i've had to modify read_env() function to allow line breaks
        - now each sequence of the backslash as a first character and
        newline as the second is replaced by two whitespaces while
        putting into the buffer

- sections where the query is constructed

        selection is made, depending on configuration variables which
        are set or not - if own query is used 





                    *-----------------------
                     2 Definitions
                    *-----------------------

#define         MAX_SUBSTITUTION_LEN    32
#define         SV_BEGIN_MARK           "$("
#define         SV_END_MARK             ")"
#define         SV_BEGIN_LEN            ((sizeof(SV_BEGIN_MARK))-1)
#define         SV_END_LEN              ((sizeof(SV_END_MARK))-1)

These definitions allows to change substitution marks in an easy way.
SV_BEGIN_MARK refers to sequence of characters treated as a prefix of
each substitution variable and SV_END_MARK refers to string which is
a closing suffix. If the expected substitution variable is called
'local_part' (without apostrophes) then '$(local_part)' is a valid
string representation for SV_BEGIN_MARK set to "$(" and SV_END_MARK to ")".
MAX_SUBSTITUTION_LEN defines maximal length of a substitution variable's
identifier (name).

The last two definitions are just for code simplification.






                    *-----------------------
                     3 New data types
                    *-----------------------

This section describes new data type definitions and variables.

3.1 struct var_data

struct var_data {                       
        const char *name;
        const char *value;
        const size_t size;
        size_t value_length;
        } ;

This structure holds information needed by parsing routines.
Using var_data array you may specify a set of string substitutions
which should be done while parsing a query. Last element in array
should have all fields set to zero (null). 

name field      - should contain substituted variable name
value           - should contain string which replaces it
size            - should contain string size including the last zero byte ('\0')
value_length    - should be set to zero - it is used as a value size cache


explanation: size is used to increase speed of calculation proccess
             value_length is used to cache length of a value during the
             parsing subroutines - it helps when substitution variable
             occures more than once within the query
 
Example:

struct var_data vdt[] = {
    {"some",    "replacement",  sizeof("some"),         0},
    {"anotha",  NULL,           sizeof("anotha"),       0},
    {NULL,      NULL,           0,                      0}
};

In this example we've declared that $(some) in the query should be
replaced by 'replacement' text, and replacement for $(anotha) will
be defined in the code before passing on the array pointer to
the paring function.


3.2 typedef size_t (*parsefunc)

typedef int (*parsefunc)(const char *, size_t, void *);

This type definition refers to the function pointer, which is used
to pass plugin functions into the core parsing subroutine. This definition
is included to simplify the declaration of the parse_core() function.





                    *-----------------------
                     4 New functions
                    *-----------------------

This section describes added functions.

4.1 get_variable

NAME

        get_variable

SYNOPSIS

        static const struct var_data *get_variable (const char *begin,
                                                    size_t len,
                                                    struct var_data *vdt);

DESCRIPTION

        This function searches an array pointed by vdt and tries to find
        the substitution variable, which name is identified with begin
        pointer and length of len bytes long.
        
        This function is also responsible for updating length cache field
        of vdt elements and validating requested variables.
        
        This function repports errors by sending human readable
        messages to the standard error stream.

RETURN VALUE

        This function returns a pointer to the array element which is
        structure of var_data type, which contains variable definition
        of a given name. It returns NULL on error or failure.


4.2 parse_core

NAME

        parse_core

SYNOPSIS
        static int    parse_core (const char *source, struct var_data *vdt,
                                  parsefunc outfn, void *result);

DESCRIPTION

        This is the parsing routine for query strings containing the
        substitution variables. It reads the string pointed with source
        and tries to catch a valid substitution variables or parts which
        are plain text blocks. The main purpose of using this function
        it to split source string into parts and for each part call
        outfn() function. Those parts are substrings identified by
        pointer to some element of the source string and size.
        Those elements are the result of splitting source string into
        logical parts: plain text substrings and substitution variables'
        values. To get the values of any found substitution variables
        parse_core() uses get_variable() function. To find places
        where substitution variables occurs it uses strstr() function
        in conjunction with SV_BEGIN_MARK and SV_END_MARK definitions.
        It passes vdt structure pointer to get_variable() function is
        it calls it.

        outfn() function should be passed by its pointer which
        refers to declaration:
        
        int (*outfn)    (const char *begin,
                         size_t string_length,
                         void *void_pointer);
        
        Each time outfn() is called the result argument of parse_core()
        is passed to the outfn() as a last argument (void_pointer).
        
        Example:
        
            Example string "$(local_part) AND $(domain)" will cause the
            outfn() to be called 3 times. First time for a value of
            $(local_part) substitution variable, second time
            for " AND " string, and the last time for $(domain) variable's
            value. Variables are passed to outfn() by their (found) values,
            plain text blocks are passed as they appear.

        This function repports errors by sending human readable
        messages to the standard error stream.

RETURN VALUE

        This function returns -1 if an error has occured and 0 if
        everything went good.
        
4.3 ParsePlugin_counter

NAME

        ParsePlugin_counter

SYNOPSIS

        int    ParsePlugin_counter (const char *begin, size_t len,
                                    void *vp);

DESCRIPTION

        This is parsing plugin function. It simply increments the value
        found in the memory area pointed by vp. It assumes that
        the memory area is allocated for the variable of size_t
        type and that area was passed by (size_t *) pointer.
        The value is incremented by len argument. Begin argument
        is not used.

        This function repports errors by sending human readable
        messages to the standard error stream.

RETURN VALUE

        This function returns the variable size or -1 if an error
        has occured, 0 if everything went good.

4.4 ParsePlugin_builder

NAME

        ParsePlugin_builder

SYNOPSIS

        int    ParsePlugin_builder (const char *begin, size_t len,
                                    void *vp);

DESCRIPTION

        This is parsing plugin function. It simply copies len bytes
        of a string pointed by begin to the end of memory area pointed by
        vp. It assumes that the area pointed by vp is passed by (char **)
        type pointer and refers to the (char *) pointer variable.
        After each call it shifts the value of pointer variable (char *)
        incrementing it by len bytes. Be careful when using this function
        - its changes the given pointer value. Always operate on an
        additional pointer type variable when passing it as the third
        argument.

RETURN VALUE

        This function returns the variable size or -1 if an error
        has occured, 0 if everything went good.

4.5 parse_string

NAME
        parse_string

SYNOPSIS

        static char *parse_string (const char *source, struct var_data *vdt);

DESCRIPTION

        This function parses the string pointed with source according to the
        replacement instructions set in var_data array, which is passed with
        its pointer vdt. It produces changed string located in newly allocated
        memory area.

        This function calls parse_core() function with various parsing
        subroutines passed as function pointers.
        
        1. It uses parse_core() with ParsePlugin_counter to obtain the
           total amount of memory needed for the output string.
        
        2. It allocates the memory.
        
        3. It uses parse_core() with ParsePlugin_builder to build the
           output string.

        This function repports errors by sending human readable
        messages to the standard error stream.

RETURN VALUE
                           
        Function returns pointer to the result buffer or NULL
        if an error has occured.

WARNINGS

        This function allocates some amount of memory using standard
        ANSI C routines. Memory allocated by this function should be
        freed with free().


4.6 validate_password

NAME
        validate_password

SYNOPSIS

        static const char *validate_password (const char *password);

DESCRIPTION

        This function checks whether password string does contain
        any dangerous characters, which may be used to pass command
        strings to the database connection stream. If it founds one
        it replaces it by the backslash character.

RETURN VALUE

        It returns a pointer to the static buffer which contains
        validated password string or NULL if an error has occured.


4.7 get_localpart

NAME
        
        get_localpart

SYNOPSIS

        static const char *get_localpart (const char *username);

DESCRIPTION

        This function detaches local part of an e-mail address
        from string pointed with username and puts it to the
        buffer of the fixed length. All necessary cleaning is
        made on the result string.

RETURN VALUE

        Pointer to the static buffer containing local part or
        NULL if there was some error.


4.8 get_domain

NAME

        get_domain

SYNOPSIS

        static const char *get_domain (const char *username, 
                                       const char *defdomain);

DESCRIPTION

        This function detaches domain part of an e-mail address
        from string pointed with username and puts it to the
        buffer of the fixed length. All necessary cleaning is
        made on the result string. If function cannot find domain
        part in the string the string pointed by defdomain is
        used instead.

RETURN VALUE

        Pointer to the static buffer containing domain name or
        NULL if there was some error.


4.9 parse_select_clause

NAME

        parse_select_clause

SYNOPSIS

        static char *parse_select_clause (const char *clause,
                                          const char *username,
                                          const char *defdomain);

DESCRIPTION

        This function is a simple wrapper to the parse_string()
        function. It parses a query pointed by caluse. username
        and defdomain strings are used to replace corresponding
        substitution strings if present in the query: $(local_part)
        and $(domain).
        

RETURN VALUE

        Same as parse_string().


4.10 parse_chpass_clause

NAME

        parse_chpass_clause

SYNOPSIS

        static char *parse_chpass_clause (const char *clause,
                                          const char *username,
                                          const char *defdomain,
                                          const char *newpass,
                                          const char *newpass_crypt);
                                                                    
DESCRIPTION

        This function is a simple wrapper to the parse_string()
        function. It parses a query pointed by caluse. username,
        defdomain, newpass and newpass_crypt strings are used to
        replace corresponding substitution strings if present in
        the query: $(local_part), $(domain), $(newpass),
        $(newpass_crypt).

RETURN VALUE

        Same as parse_string().





                    *------------------------
                     5 Ideas and TODO
                    *------------------------

- solve problem with fixed buffer length of local part and the domain part
  strings after split                                           (problem?)
- allow admin to set a group name instead of numerical group id
- allow admin to set a username instead of numerical user id

- add clauses:

  - MYSQL_PRESELECT_CLAUSE (query which comes before MYSQL_SELECT_CLAUSE)
  - MYSQL_POSTSELECT_CLAUSE (query which comes after MYSQL_SELECT_CLAUSE)
  - PGSQL_PRESELECT_CLAUSE (query which comes before PGSQL_SELECT_CLAUSE)
  - PGSQL_POSTSELECT_CLAUSE (query which comes after PGSQL_SELECT_CLAUSE)





                    *------------------------
                     6 Thanks
                    *------------------------

At the beginning this patch was messy indeed. :> I would like to thank
Sam Varshavchik for pointing me a lot how to make it more fast and solid.
I would also thank Philip Hazel, Chris Lightfoot and Mike Bremford which
by their software capabilities inspired me to write it.

Thomas T. Thai <tom@minnesota.com> ported author's original MySQL code
to the PostgreSQL module.

---------------------------------------------------------------------------