Data Structure for Multi-Level Traversal

I have an application that deals with data in the following structure:

struct Message
{
   int    time;
   string name;
   string details;
};

For example, I may have a data set that looks like the following:

9:00:00 Bob  <Info>
9:01:00 John <Info>
9:05:00 Bob  <Info>
9:11:00 Mary <Info>
9:17:00 John <Info>
9:25:00 Mary <Info>
9:30:00 Bob  <Info>

And I will have a list of Message structures that represent each line in the data set.

Some operations I will need to do on this data include:

• Collect all data in chronological order and dostuff()
• Collect all data from John (or whoever) in chronological order and dostuff()

So, I need a way to traverse the list such that I can pass every messag开发者_开发问答e in chronological order, and also choose a person, and pass through only their messages in chronological order.

My thoughts are to have a struct like this:

struct Node
{
   Message* message;
   Node*    next_time;
   Node*    next_name;
};

In which next_time points to the next Node in chronological order, and next_name points to the next Node that belongs to message->name. And a Root structure points to the first of each type.

struct Root 
{
   Node* first_time;
   Node* first_bob;
   Node* first_john;
   Node* first_mary;

   Node* last_time;
   Node* last_bob;
   Node* last_john;
   Node* last_mary;
};

Here is an image to illustrate the point.

This structure allows me to fairly easily traverse through every message, or through only Bob's messages, or only John's, etc.

However, I am worried that maybe this is more complicated than it needs to be. I also have concerns about maintenance (see below). I need the search/select/read operations to be pretty fast, which I think they are. And I need insert operations to be reasonably fast. But right now, for every Message I insert, I must (1) update some next_time pointer and (2) update some next_name pointer.

My question is: Does a data structure exist that provides this type of functionality already? If not, am I approaching this problem correctly?

Please provide any code samples in C++ or C#, if possible.

Thanks.

Additional: Suppose later I want to add to my Message struct. Let's say I add a field called City. Now, I may want to do this:

• Collect all data from a specific City in chronological order and dostuff()

This would require adding a next_city, and then for every insert, I would have to update next_time, next_name, AND next_city.

Further, suppose I want to do this:

• Collect all data from a specific City AND a specific name in chronological order and dostuff()

I think this makes the problem incredibly more difficult unless I opt to traverse every Message and skip the ones I don't care about.

---

A simple linked list of all messages (sorted by time).

struct Node
{
   Message* message;
   Node*    next_name;
};

This will satisfy req 1. You can add() and getAll() in O(1)

A separate hashmap with User as key and a list of Node* as values.

Hashmap
{key = User, value = List(Message*)}

This will satisfy req 2. You can add a new entry to the end of the list of the specific user O(1) and getAllOfUser() can also run in O(1)

---

I would probably create a class to represent a user, storing users in an hash table by some identifier like the name, then have each user hold a list of Messages sorted in chronological order, which are also stored in a single global list which holds everyone's Messages in chronological order. For every added Message, you'd have to insert it once in each list (by list I mean some collection), which could be log n time, or as bad as n, depending on the data structure.