Lamac big execution time

[merge34]

Hello. During my progress in the compiler course I've encountered a possible bug in lamac.

Scenario

1. Add case ... of construction to the code
2. Add a considerable number (10+) of string patterns
3. Add function call using dot notation after ->
4. Run lamac <file name>

Problem In the described scenario lamac execution time grows exponentially as the number of patterns grows.

Expected behavior Execution time in the described scenario is similar to the execution time of lamac without dot notation usage.

Example Consider the following gist, in which I've implemented 2 semantically equivalent programs, according to specification:

1. In file example1.lama I've implemented the described scenario without dot notation usage. So, without any dependence on N, we have something like:

   $ time lamac test.lama
   real    0m8,152s                                                                                                                   
   user    0m8,037s  
   sys     0m0,102s

2. In file example2.lama I've implemented exactly described scenario. So, we have:

• N = 0

  $ time lamac test.lama
  real    0m0,227s  
  user    0m0,219s 
  sys     0m0,009s

• N = 3

  $ time lamac test.lama          
  real    0m4,357s                    
  user    0m4,289s                   
  sys     0m0,067s

• N = 5

  $ time lamac test.lama 
  real    1m23,222s 
  user    1m22,964s
  sys     0m0,223s

• N = 7 Did not finish after 10 minutes of execution

Environment description

• Ubuntu 18.04
• Opam 2.0.5
• OCaml 4.0.5
• gcc-multilib 7.4.0

[dboulytchev]

Looks like Ostap issue with dot notation support. Should be postponed until hash-consing and proper memoization is implemented.