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

# ## Analyze access color trends using Unpaywall data

# In[1]:


import collections

import pandas
import plotnine
import numpy

import utils


# In[2]:


access_df = pandas.read_csv(
    'data/02.unpaywall-access.tsv.xz',
    sep='\t',
    #nrows=100_000,
)
access_df = access_df.query("crossref_type in @utils.get_crossref_types()")
access_df['journal_year'] = access_df.journal_date.str.slice(stop=4).astype('float').astype('Int64')
access_df.head(2)


# In[3]:


access_df['journal_access_category'] = 'closed'
access_df.loc[access_df.unpaywall_access.astype(bool), 'journal_access_category'] = 'green'
access_df.loc[access_df.journal_access.astype(bool), 'journal_access_category'] = 'hybrid/bronze'
access_df.loc[access_df.journal_fully_oa.astype(bool), 'journal_access_category'] = 'gold'
access_df.head()


# In[4]:


access_df.journal_access_category.value_counts()


# In[5]:


access_df.crossref_type.value_counts()


# In[6]:


def summarize(df):
    row = collections.Counter()
    counts = collections.Counter(df.journal_access_category)
    for category in utils.access_categories_colors:
        row[category] = counts[category]
    row['total'] = len(df)
    return pandas.Series(row)

year_df = access_df.groupby('journal_year').apply(summarize).reset_index()
year_df.journal_year = year_df.journal_year.astype(int)
year_df = year_df.query("1950 <= journal_year <= 2018")
year_df.tail(10)


# In[7]:


year_df.to_csv('data/05.unpaywall-colors-by-year.tsv', sep='\t', index=False)


# In[9]:


year_melt_df = year_df.melt(id_vars=['journal_year', 'total'], var_name='access_category', value_name='count')
year_melt_df.access_category = pandas.Categorical(
    year_melt_df.access_category,
    categories=reversed(list(utils.access_categories_colors))
)
year_melt_df.head(2)


# In[10]:


gg_counts = (
    plotnine.ggplot(mapping=plotnine.aes(x='journal_year', y='count', fill='access_category'), data=year_melt_df) +
    plotnine.geom_bar(stat='identity', size=0, color=None, width=1) +
    plotnine.scale_x_continuous(breaks=numpy.arange(1960, 2011, 10), expand=(0, 0), name='') +
    plotnine.scale_y_continuous(labels=lambda array: [f'{x:,.0f}' for x in array], name='', expand=(0,0)) +
    plotnine.scale_fill_manual(values=utils.access_categories_colors) +
    plotnine.theme_bw() +
    plotnine.theme(legend_position=(0.3, 0.7), legend_title=plotnine.element_blank())
)
gg_counts


# In[12]:


gg_percents = (
    plotnine.ggplot(mapping=plotnine.aes(x='journal_year', y='count', fill='access_category'), data=year_melt_df) +
    plotnine.geom_bar(stat='identity', position='fill', size=0, color=None, width=1) +
    plotnine.scale_fill_manual(values=utils.access_categories_colors) +
    plotnine.scale_x_continuous(breaks=numpy.arange(1960, 2011, 10), expand=(0, 0), name='') +
    plotnine.scale_y_continuous(breaks=numpy.arange(0.1, 1, 0.2), labels=lambda array: [f'{x:.0%}' for x in array], expand=(0, 0), name='') +
    plotnine.theme_bw() +
    plotnine.theme(legend_position=(0.3, 0.7), legend_title=plotnine.element_blank())
)
gg_percents