#!/usr/bin/env python
# coding: utf-8

# # Function arities of different Scala projects
# 
# I was curious what is a distribution of function arities in different Scala projects.
# Scala is a functional programming language, so I assumed the average number of arguments would be more than 2.
# 
# So I did a small research. I created a Scala compiler plugin that gathered defined functions arities.
# In this notebook I visualize collected data for the following projects:
# - Scala 2.12.4 compiler and standard library
# - Dotty 0.5.0-RC1 compiler
# - Akka 2.4
# - Spark 2.2
# - Cats 1.0
# - Scalaz 7.2.9

# In[10]:


get_ipython().run_line_magic('matplotlib', 'inline')
import matplotlib.pyplot as plt  
import numpy as np


# I don't count constructors, and do _not_ count `this` as implicit argument

# In[11]:


def readStats(filename):
    x = []
    with open(filename, "r") as file:    
        for l in file:
            name, num, implicits = l.strip().split()
            n = int(num)
            if name == "<init>":
                continue    
            x.append(n)
    return np.asarray(x)


# In[12]:


scalaz = readStats("data/scalaz8.txt")
akka   = readStats("data/akka2.4.txt")
scala  = readStats("data/scala2.12.txt")
dotty  = readStats("data/dotty.txt")
spark  = readStats("data/spark2.2.txt")
cats   = readStats("data/cats1.0.txt")


# In[13]:


plt.style.use('seaborn')
fig = plt.figure(figsize=(12, 5))
fig.suptitle("Function arities in Scala projects")
plt.subplot(121)
plt.title("Scala 2.12")
plt.text(4, 8000, "mean: %.2f  std: %.2f" % (np.mean(scala), np.std(scala)))
n, bins, patches = plt.hist(scala, bins=range(8),  alpha=0.6, histtype='bar')
plt.subplot(122)
plt.title("Dotty 0.5.0-RC1")
plt.text(4, 2500, "mean: %.2f  std: %.2f" % (np.mean(dotty), np.std(dotty)))
n, bins, patches = plt.hist(dotty, bins=range(8),  alpha=0.6, histtype='bar')


# In[14]:


fig = plt.figure(figsize=(12, 5))
fig.suptitle("Function arities in Scala projects")
plt.subplot(121)
plt.title("Akka 2.4")
plt.text(4, 10000, "mean: %.2f  std: %.2f" % (np.mean(akka), np.std(akka)))
n, bins, patches = plt.hist(akka, bins=range(8),  alpha=0.6, histtype='bar')
plt.subplot(122)
plt.title("Spark 2.2")
plt.text(4, 8000, "mean: %.2f  std: %.2f" % (np.mean(spark), np.std(spark)))
n, bins, patches = plt.hist(spark, bins=range(8),  alpha=0.6, histtype='bar')


# In[16]:


fig = plt.figure(figsize=(12, 5))
fig.suptitle("Function arities in Scala projects")
plt.subplot(121)
plt.title("Cats 1.0")
plt.text(4, 800, "mean: %.2f  std: %.2f" % (np.mean(cats), np.std(cats)))
n, bins, patches = plt.hist(cats, bins=range(8),  alpha=0.6, histtype='bar')
plt.subplot(122)
plt.title("ScalaZ 8")
plt.text(4, 250, "mean: %.2f  std: %.2f" % (np.mean(scalaz), np.std(scalaz)))
n, bins, patches = plt.hist(scalaz, bins=range(8),  alpha=0.6, histtype='bar')


# # Conclusions
# 
# There are no conclusions. Just a general 'Aha, that's how it is'. I was surprised by the number of zero argument functions.

# # Future work
# 
# I'd like to gather similar statistics for Haskell projects.
# 
# Please, contribute!