Add option to store all world states in information state tree

[maichmueller]

This PR adds a boolean flag to the constructor of the infostate tree to allow storing the corresponding world states and chance reach probability not just at the leaf nodes, but all infostate nodes (except filler nodes).

In particular, this change allows to attain an inverse mapping of info state to world states when traversing a player's infostate tree.

The changes are:

• added boolean flag to InfostateTree class
• change logic in building tree to account for the flag
• adopt info state tree tests to account for the change
• added is_filler_node() check to InfostateNode class to query node
• extended child_at method of InfostateNode to choose to skip filler nodes via bool flag
• exchange const std::string& with std::string_view in a couple of appropriate places in infostate_tree.(h|cpp)

An example output of traversing player 0's infostate tree of 2-player Kuhn Poker (skipping filler nodes):

Child-Sequence: []                           type: observation     #children: 1     states: []                   chance_reach_prob: []                       
Child-Sequence: [0]                          type: observation     #children: 3     states: []                   chance_reach_prob: ['1.000']                
Child-Sequence: [0, 2]                       type: observation     #children: 1     states: [2]                  chance_reach_prob: ['0.333']                
Child-Sequence: [0, 2, 0]                    type: decision        #children: 2     states: [2 0, 2 1]           chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 2, 0, 1]                 type: observation     #children: 1     states: [2 0 b, 2 1 b]       chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 2, 0, 1, 0]              type: observation     #children: 4     states: [2 0 b, 2 1 b]       chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 2, 0, 1, 0, 3]           type: terminal        #children: 0     states: [2 1 bb]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 2, 0, 1, 0, 2]           type: terminal        #children: 0     states: [2 1 bp]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 2, 0, 1, 0, 1]           type: terminal        #children: 0     states: [2 0 bb]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 2, 0, 1, 0, 0]           type: terminal        #children: 0     states: [2 0 bp]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 2, 0, 0]                 type: observation     #children: 1     states: [2 0 p, 2 1 p]       chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 2, 0, 0, 0]              type: observation     #children: 3     states: [2 0 p, 2 1 p]       chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 2, 0, 0, 0, 2]           type: terminal        #children: 0     states: [2 1 pp]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 2, 0, 0, 0, 1]           type: decision        #children: 2     states: [2 0 pb, 2 1 pb]     chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 2, 0, 0, 0, 1, 1]        type: observation     #children: 2     states: [2 0 pbb, 2 1 pbb]   chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 2, 0, 0, 0, 1, 1, 1]     type: terminal        #children: 0     states: [2 1 pbb]            chance_reach_prob: ['0.167']                
Child-Sequence: [0, 2, 0, 0, 0, 1, 1, 0]     type: terminal        #children: 0     states: [2 0 pbb]            chance_reach_prob: ['0.167']                
Child-Sequence: [0, 2, 0, 0, 0, 1, 0]        type: observation     #children: 2     states: [2 0 pbp, 2 1 pbp]   chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 2, 0, 0, 0, 1, 0, 1]     type: terminal        #children: 0     states: [2 1 pbp]            chance_reach_prob: ['0.167']                
Child-Sequence: [0, 2, 0, 0, 0, 1, 0, 0]     type: terminal        #children: 0     states: [2 0 pbp]            chance_reach_prob: ['0.167']                
Child-Sequence: [0, 2, 0, 0, 0, 0]           type: terminal        #children: 0     states: [2 0 pp]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 1]                       type: observation     #children: 1     states: [1]                  chance_reach_prob: ['0.333']                
Child-Sequence: [0, 1, 0]                    type: decision        #children: 2     states: [1 0, 1 2]           chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 1, 0, 1]                 type: observation     #children: 1     states: [1 0 b, 1 2 b]       chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 1, 0, 1, 0]              type: observation     #children: 4     states: [1 0 b, 1 2 b]       chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 1, 0, 1, 0, 3]           type: terminal        #children: 0     states: [1 2 bb]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 1, 0, 1, 0, 2]           type: terminal        #children: 0     states: [1 2 bp]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 1, 0, 1, 0, 1]           type: terminal        #children: 0     states: [1 0 bb]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 1, 0, 1, 0, 0]           type: terminal        #children: 0     states: [1 0 bp]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 1, 0, 0]                 type: observation     #children: 1     states: [1 0 p, 1 2 p]       chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 1, 0, 0, 0]              type: observation     #children: 3     states: [1 0 p, 1 2 p]       chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 1, 0, 0, 0, 2]           type: terminal        #children: 0     states: [1 2 pp]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 1, 0, 0, 0, 1]           type: decision        #children: 2     states: [1 0 pb, 1 2 pb]     chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 1, 0, 0, 0, 1, 1]        type: observation     #children: 2     states: [1 0 pbb, 1 2 pbb]   chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 1, 0, 0, 0, 1, 1, 1]     type: terminal        #children: 0     states: [1 2 pbb]            chance_reach_prob: ['0.167']                
Child-Sequence: [0, 1, 0, 0, 0, 1, 1, 0]     type: terminal        #children: 0     states: [1 0 pbb]            chance_reach_prob: ['0.167']                
Child-Sequence: [0, 1, 0, 0, 0, 1, 0]        type: observation     #children: 2     states: [1 0 pbp, 1 2 pbp]   chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 1, 0, 0, 0, 1, 0, 1]     type: terminal        #children: 0     states: [1 2 pbp]            chance_reach_prob: ['0.167']                
Child-Sequence: [0, 1, 0, 0, 0, 1, 0, 0]     type: terminal        #children: 0     states: [1 0 pbp]            chance_reach_prob: ['0.167']                
Child-Sequence: [0, 1, 0, 0, 0, 0]           type: terminal        #children: 0     states: [1 0 pp]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 0]                       type: observation     #children: 1     states: [0]                  chance_reach_prob: ['0.333']                
Child-Sequence: [0, 0, 0]                    type: decision        #children: 2     states: [0 1, 0 2]           chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 0, 0, 1]                 type: observation     #children: 1     states: [0 1 b, 0 2 b]       chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 0, 0, 1, 0]              type: observation     #children: 4     states: [0 1 b, 0 2 b]       chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 0, 0, 1, 0, 3]           type: terminal        #children: 0     states: [0 2 bb]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 0, 0, 1, 0, 2]           type: terminal        #children: 0     states: [0 2 bp]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 0, 0, 1, 0, 1]           type: terminal        #children: 0     states: [0 1 bb]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 0, 0, 1, 0, 0]           type: terminal        #children: 0     states: [0 1 bp]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 0, 0, 0]                 type: observation     #children: 1     states: [0 1 p, 0 2 p]       chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 0, 0, 0, 0]              type: observation     #children: 3     states: [0 1 p, 0 2 p]       chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 0, 0, 0, 0, 2]           type: terminal        #children: 0     states: [0 2 pp]             chance_reach_prob: ['0.167']                
Child-Sequence: [0, 0, 0, 0, 0, 1]           type: decision        #children: 2     states: [0 1 pb, 0 2 pb]     chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 0, 0, 0, 0, 1, 1]        type: observation     #children: 2     states: [0 1 pbb, 0 2 pbb]   chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 0, 0, 0, 0, 1, 1, 1]     type: terminal        #children: 0     states: [0 2 pbb]            chance_reach_prob: ['0.167']                
Child-Sequence: [0, 0, 0, 0, 0, 1, 1, 0]     type: terminal        #children: 0     states: [0 1 pbb]            chance_reach_prob: ['0.167']                
Child-Sequence: [0, 0, 0, 0, 0, 1, 0]        type: observation     #children: 2     states: [0 1 pbp, 0 2 pbp]   chance_reach_prob: ['0.167', '0.167']       
Child-Sequence: [0, 0, 0, 0, 0, 1, 0, 1]     type: terminal        #children: 0     states: [0 2 pbp]            chance_reach_prob: ['0.167']                
Child-Sequence: [0, 0, 0, 0, 0, 1, 0, 0]     type: terminal        #children: 0     states: [0 1 pbp]            chance_reach_prob: ['0.167']                
Child-Sequence: [0, 0, 0, 0, 0, 0]           type: terminal        #children: 0     states: [0 1 pp]             chance_reach_prob: ['0.167']                

[maichmueller]

Hi @lanctot, when in Rome of old PRs: what is your view on this one?

[lanctot]

Ah yes, good. Straight-forward, should be doable. Can you also rebase to make it current, if you haven't already?

[maichmueller]

should this be merged, i would need to update the bindings in #1054 to expose the feature as well

[lanctot]

Ok let's do this one first and you can then update #1054 with a pull from master once it's in.

[lanctot]

Here is an example of things we'd still need to catch in #1071 found when importing this PR.

• Lines should be <= 80 columns wide
• Missing space before ( in if(
• Use operator || instead of or
• Use operator && instead of and
• Use operator ! instead of not
• Missing include for DCHECK (spiel_utils.h)
• Missing header for size_t (cstddef)
• Missing header for string