readme

brainstorming

features

• inspiration: networkx api

• documentation

  • algo/derivations + write up thingy (data - "we chose this graph representation based on these benchmarks we did: xd")

• ease of use

  • simple algos correctly & quickly

• header-only

• flexibility/parameterization

• robust (cycles, negative edge weights, invalid inputs, etc)

• examples - performance comparisons (boost, )

• integrate with STL/STL-like api

  • iterators

• package manager integration

• easy to define ur own graph algorithms using library

  • modular (i.e. do custom shit)

• big objects

• multithreading (?)/parallel processing

  • thread-safe graphs

• niche shit

  • compression

/*
    std::vector::{reverse,bidirectional,forward}_iterator
    std::graph::{topo,dfs,bfs}_iterator
    std::tree::{level}_iteratir <- is a bfs iterator - could maybe typedef it for clarity
    - same w/ dfs as preorder - ways to find next? flexibility...
*/

algorithms

• native operator support
• flexible comparators/lambdas and stuff
  • all compile-time validated
• also: TREES
  • proper polymorphism between calls (i.e. returning tree from graph::{*})
  • ^ make compile-time as much as possible

graph<node>([](const node n) { return n.value; })

// spam parameters -> graph(a, b, c); what if you just want to specify c?
//  - no kwargs in c++
// pass config: graph(config)
// certain specalized classes:


// tree w/ tree-specific algos/apis
// disable certain graph shits
// enable others: level_order
class tree : graph {
    T root;
};

class graph {
    bool directed;
    bool weighted;
    double sparsity();
    bool cyclical();
};

class telegraph {
    graph mst();
};

infra

• make
• testing + benchmarking
• documentation (non-dogshit)
  • intellisense-like signature help
  • NOT doxygen