std::map proper use?

[bkelly]

Environment Windows XP Pro, Visual Studio, C++, MFC
After examining some code I use to put data in a map I have become concerned that I may be inserting data improperly. The map for parameters is defined as:

struct td_parameter_definitions
{
CString          name;
unsigned ing  tag;
...
};

typedef std::map< td_parameter_key, td_parameter_definitions > td_parameter_map;

The parameter map is within a message structure. The address of the message structure is handed to a method that reads a text file and fills in data. It creates a local pointer to the parameter map and sets fields within the parameter structure as follows:

Read_parameter_Definitions( st_msg_data *p_msg_data )
{ ...
   // make a short hand name for the parameter map
td_parameter_map *p_parm_map = p_msg_data->parameter_map;
parameter_number = <an integer value>;
...
       // cs_t1 is filled in by some additional code not shown here.
( *p_parm_map)[parameter_number].name = cs_t1; 
// etc
... }

My problem is that I realized that I was not explicitly adding entry [parameter_number] to the map. The code just jumped up and starting putting data in that entry as though it already exists. I am finding that rather worrying.

Since the map is defined as a map of structures and not defined as a map of pointers, does the simple act of referencing entry [1] (for example) create that entry. If so, then is it OK for the code to proceed to set various fields of that entry.

Is that a back door to creating an entry in a map that I should not use?

[mattpd]

Yes, it's standard behavior, operator[] of std::map inserts elements if they don't exist, it's intentional:
http://www.cplusplus.com/reference/s...tor%5B%5D/

If you merely want to check for existence, use the "find" member-function:
http://www.cplusplus.com/reference/stl/map/find/
or, alternatively, "count":
http://www.cplusplus.com/reference/stl/map/count/

Note that for std::multimap using "find" would be preferred (since std::multimap allows for duplicate elements its "count" mem.-fn. would have more work to do), but for std::map (which doesn't) I doubt the choice matters.

[bkelly]

Cool, thanks for the reply.

[monarch_dodra]

In "theory" the "best" method is using a "lower_bound", and checking the returned iterator. The advantage to this is that if it does not find your element, it still returns an "hint" for a more efficient future insertion.

That said, this should be reserved for the "optimization domain". IMO, i've always used operator[]. I try to prefer code that talks to me.

[D_Drmmr]

My problem is that I realized that I was not explicitly adding entry [parameter_number] to the map. The code just jumped up and starting putting data in that entry as though it already exists. I am finding that rather worrying.

Since the map is defined as a map of structures and not defined as a map of pointers, does the simple act of referencing entry [1] (for example) create that entry. If so, then is it OK for the code to proceed to set various fields of that entry.

Whether this is ok depends on whether it is ok to overwrite an existing entry in the map. If it is, then the code is fine. It it's not, but the choice of your key should prevent this, then make sure you add an assertion that checks that the key is not present before using the operator []. If it's possible that the key is already present in the map and it should not be overwritten, then you'll have to write more verbose code using a combination of find/lower_bound and insert.

[JohnW@Wessex]

Just a minor style point, but you could make your code a little easier to read by using references rather than pointers.

td_parameter_map &parm_map = *p_msg_data->parameter_map; 
parm_map[parameter_number].name = cs_t1;

This is especially relevent if the pointer is not expected to be changed.