endpoints.py

"""Functions linked directly to functionality called from API endpoints.

"""
from datetime import datetime

import bottle

import config
import db
import helpers
import models

def get_categories(connection,repo='all'):
  """Fetches a list of all known preprint categories.

  bioRxiv separates all papers into categories (or "collections"), such
  as "bioinformatics", "genomics", etc. This function lists all the ones
  we've pulled from the site so far.

  Arguments:
    - connection: a Connection object with an active database session

  Returns:
    - a list of strings, one for each collection

  """
  results = []
  query = """
    SELECT DISTINCT collection
    FROM articles
    WHERE collection IS NOT NULL
  """
  if repo != 'all':
    query += " AND repo=%s"
    params = (repo,)
  else:
    params = ()
  query += " ORDER BY collection"
  categories = connection.read(query, params)
  for cat in categories:
    if len(cat) > 0:
      results.append(cat[0])
  return results

def paper_query(q, categories, timeframe, metric, page, page_size, repo, version, connection):
  """Returns a list of the most downloaded papers that meet a given set of constraints.

  Arguments:
    - connection: a database Connection object.
    - q:  A search string to compare against article abstracts,
          titles and author names. (Title matches are weighted more heavily.)
    - categories: A list of categories the results can be in.
    - timeframe: A description of the range of dates on which to
          base the rankings (i.e. "alltime" or "lastmonth")
    - metric: Which article-level statistic to use when sorting results
    - page: Which page of the results to display (0-indexed)
    - page_size: How many entries should be returned
  Returns:
    - An list of Article objects that meet the search criteria, sorted by the
          specified metric in descending order.

  """

  # HACK: Because there are so many possible combinations for which
  # parameters need to be passed to the database for this query,
  # it's much easier to say that every query needs a "repo" parameter
  # rather than having lots of nested options in which some of them
  # are performed without any repository clause.
  if repo == 'all':
    repo = ['biorxiv','medrxiv']
  else:
    repo = [repo]


  # We build two queries, 'select' and 'countselect': one to get the
  # current page of results, and one to figure out the total number
  # of results
  select = "SELECT "
  if metric == "downloads":
    select += "r.downloads"
  elif metric == "twitter":
    select += "SUM(r.count)"
  select += ", a.id, a.url, a.title, a.abstract, a.collection, a.posted, a.doi, a.repo"

  countselect = "SELECT COUNT(DISTINCT a.id)"
  params = (repo,)

  query = ""
  if q != "": # if there's a text search specified
    params = (q,repo)
  query += f' FROM {config.db["schema"]}.articles AS a INNER JOIN {config.db["schema"]}.'
  if metric == "twitter":
    query += "crossref_daily"
  elif metric == "downloads":
    query_times = {
      "alltime": "alltime_ranks",
      "ytd": "ytd_ranks",
      "lastmonth": "month_ranks",
    }
    query += query_times[timeframe]

  if metric == "twitter":
    query += " AS r ON r.doi=a.doi"
  elif metric == "downloads":
    query += " AS r ON r.article=a.id"

  if q != "":
    # backwards compatibility for text search
    if version == 1:
      query += """, plainto_tsquery(%s) query,
      coalesce(setweight(a.title_vector, 'A') || setweight(a.abstract_vector, 'C') || setweight(a.author_vector, 'D')) totalvector
      """
    else:
      query += """, websearch_to_tsquery(%s) query,
      coalesce(setweight(a.title_vector, 'A') || setweight(a.abstract_vector, 'C') || setweight(a.author_vector, 'D')) totalvector
      """
  # add a WHERE clause if we need one:
  # (all-time twitter stats don't require it)
  query += f" WHERE a.repo=ANY(%s)"
  if metric == "downloads" or (metric == "twitter" and timeframe != "alltime") or len(categories) > 0:
    query += " AND "
    if metric == "downloads":
      query += "r.downloads > 0"
      if q != "" or len(categories) > 0:
        query += " AND "
    if q != "":
      query += "query @@ totalvector "
      if len(categories) > 0 or (metric == "twitter" and timeframe != "alltime"):
        query += " AND "

    if len(categories) > 0:
      query += "collection=ANY(%s)"
      if q != "":
        params = (q,repo,categories)
      else:
        params = (repo,categories)
      if metric == "twitter" and timeframe != "alltime":
        query += " AND "
    if metric == "twitter" and timeframe != "alltime":
      query += "r.source_date > now() - interval "
      query_times = {
        "day": 2,
        "week": 7,
        "month": 30,
        "year": 365
      }
      query += f"'{query_times[timeframe]} days' "
  # this is the last piece of the query we need for the one
  # that counts the total number of results
  countselect += query
  resp = connection.read(countselect, params)
  total = resp[0][0]
  # continue building the query to get the full list of results:
  if metric == "twitter":
    query += " GROUP BY a.id"
  query += " ORDER BY "
  if metric == "downloads":
    query += "r.rank ASC"
  elif metric == "twitter":
    query += "SUM(r.count) DESC"

  query += f" LIMIT {page_size}"
  if page > 0:
    query += f" OFFSET {page * page_size}"
  query += ";"

  select += query
  result = connection.read(select, params)
  results = [models.SearchResultArticle(a, connection) for a in result]
  return results, total

def author_rankings(connection, category=""):
  """Fetches a list of authors with the most cumulative downloads.

  Arguments:
    - connection: a database Connection object.
    - category: (Optionally) a single collection to base download rankings on.
  Returns:
    - A list of Author objects that meet the search criteria.

  """
  if category == "": # all time, all categories
    table = "author_ranks"
    where = ""
    params = ()
  else:
    table = "author_ranks_category"
    where = "WHERE r.category=%s"
    params = (category,)
  query = f"""
    SELECT a.id, a.name, r.rank, r.downloads, r.tie
    FROM authors AS a
    INNER JOIN {config.db["schema"]}.{table} r ON a.id=r.author
    {where}
    ORDER BY r.rank
    LIMIT {config.author_ranks_limit}
  """

  authors = connection.read(query, params)
  return [models.SearchResultAuthor(*a) for a in authors]

def author_details(author_id, connection):
  """Returns information about a single author, including a list of
      all their papers.

  Arguments:
    - connection: a database Connection object.
    - author_id: the Rxivist-issued ID of the author being queried.
  Returns:
    - An Author object containing information about that
        author's publications and contact info.

  """
  result = models.Author(author_id)
  result.GetInfo(connection)
  return result

def paper_details(article_id, connection):
  """Returns information about a single paper.

  Arguments:
    - connection: a database Connection object.
    - article_id: the ID given to the author being queried.
  Returns:
    - A Paper object containing details about the paper and
        its authors.

  """
  result = models.ArticleDetails(article_id, connection)
  return result

def paper_downloads(a_id, connection):
  """Returns time-series data about how many
  times a paper's webpage and PDF have been downloaded.

  Arguments:
    - connection: a database Connection object.
    - a_id: the Rxivist-issued ID given to the paper being queried.
  Returns:
    - A list of months and the download stats for each month

  """
  result = models.Article(a_id)
  result.GetTraffic(connection)
  return {
    "query": {
      "id": a_id
    },
    "results": [{"month": x.month, "year": x.year, "downloads": x.downloads, "views": x.views} for x in result.traffic]
  }

def get_distribution(category, metric, connection):
  """Returns time-series data about how many
  times a paper's webpage and PDF have been downloaded.

  Arguments:
    - connection: a database Connection object.
    - a_id: the Rxivist-issued ID given to the paper being queried.
  Returns:
    - A list of months and the download stats for each month

  """
  # "category" param can be either "author" or "paper"
  # "metric" param is (right now) limited to just "downloads"

  if category == "paper":
    category = "alltime"
  data = connection.read("SELECT bucket, count FROM download_distribution WHERE category=%s ORDER BY bucket", (category,))
  results = [(entry[0], entry[1]) for entry in data]
  averages = {}
  for avg in ["mean", "median"]:
    cat = f"{category}_{avg}"
    answer = connection.read("SELECT count FROM download_distribution WHERE category=%s", (cat,))
    averages[avg] = answer[0][0]
  return results, averages

def top_year(year, connection):
  resp = connection.read("""
  SELECT SUM(t.pdf) as downloads, t.article, a.url,
    a.title, a.abstract, a.collection, a.posted, a.doi
    FROM article_traffic t
    INNER JOIN articles a ON t.article=a.id
    WHERE t.year = %s
      AND a.posted >= '%s-01-01'
      AND a.posted <= '%s-12-31'
    GROUP BY 2,3,4,5,6,7,8
    ORDER BY 1 DESC
    LIMIT 25
  """, (year,year,year))
  if len(resp) == 0:
    return []
  results = [models.SearchResultArticle(a, connection) for a in resp]
  return results

def summary_stats(connection, category=None):
  """Returns time-series data reflecting how many submissions and downloads
  were recorded in each month.

  Arguments:
    - connection: a database Connection object.
    - category: *tk NOOP whether to restrict the query to only a single collection.
  Returns:
    - A dictionary containing collections, each of which stores either the submission
        or download numbers for a single month.

  """
  results = {
    'submissions': [],
    'downloads': []
  }

  # Submissions:
  # The reason this is so complicated is because bioRxiv has more
  # months of submissions, but we want an entry for each month.

  # figure out the most recent month, so we know when to stop:
  data = connection.read(f"SELECT MAX(EXTRACT(YEAR FROM posted)) FROM {config.db['schema']}.articles")
  maxyear = int(data[0][0])
  data = connection.read(f"SELECT MAX(EXTRACT(MONTH FROM posted)) FROM {config.db['schema']}.articles WHERE EXTRACT(YEAR FROM posted) = %s", (maxyear,))
  maxmonth = int(data[0][0])

  for repo in ['biorxiv','medrxiv']:
    repodata = {}
    for year in range(2013, maxyear + 1):
      repodata[year] = {}
      for month in range(1,13):
        if year == 2013 and month < 11:
          continue # nothing before nov 2013
        if year == maxyear and month > maxmonth:
          break
        repodata[year][month] = 0
    data = connection.read("""
      SELECT EXTRACT(MONTH FROM posted)::int AS month,
        EXTRACT(YEAR FROM posted)::int AS year, COUNT(id) AS submissions
      FROM prod.articles
      WHERE posted IS NOT NULL
      AND repo=%s
      GROUP BY year, month
      ORDER BY year, month;
    """,(repo,))
    for entry in data:
      repodata[entry[1]][entry[0]] = entry[2]

    monthdata = []
    for year, yeardata in repodata.items():
      for month, count in yeardata.items():
        monthdata.append({
          'month': month,
          'year': year,
          'count': count,
        })
    results['submissions'].append({
      'label': repo,
      'data': monthdata
    })

  # Downloads:

  # Don't show the previous month's download numbers until X days after the
  # month ends, since those numbers won't be updated for every paper until
  # after the month ends
  current = datetime.now().day
  adjust = 1 if current > config.summary_download_age else 2

  if maxmonth > adjust:
    maxmonth -= adjust
  else:
    maxmonth += 12 - adjust
    maxyear -= 1

  for repo in ['biorxiv','medrxiv']:
    repodata = {}
    for year in range(2013, maxyear + 1):
      repodata[year] = {}
      for month in range(1,13):
        if year == 2013 and month < 11:
          continue # nothing before nov 2013
        if year == maxyear and month > maxmonth:
          break
        repodata[year][month] = 0
    data = connection.read("""
      SELECT t.month, t.year, sum(t.pdf) AS downloads
      FROM prod.article_traffic t
      INNER JOIN prod.articles a ON t.article=a.id
      WHERE repo=%s
      GROUP BY year, month
      ORDER BY year, month
    """,(repo,))
    for entry in data:
      # skip results outside the range we want
      if entry[1] > maxyear:
        continue
      if entry[1] == maxyear and entry[0] > maxmonth:
        continue
      repodata[entry[1]][entry[0]] = entry[2]

    monthdata = []
    for year, yeardata in repodata.items():
      for month, count in yeardata.items():
        monthdata.append({
          'month': month,
          'year': year,
          'count': count,
        })
        if year == maxyear and month == maxmonth:
          break

    results['downloads'].append({
      'label': repo,
      'data': monthdata
    })

  return results

def site_stats(connection):
  """Returns a (very) brief summary of the information indexed by Rxivist. More of
  a data hygiene report than anything.

  Arguments:
    - connection: a database Connection object.
  Returns:
    - A dict with the total indexed papers and authors

  """

  # Counting up how many of each entity we have
  resp = connection.read("SELECT COUNT(id) FROM articles;")
  if len(resp) != 1 or len(resp[0]) != 1:
    paper_count = 0
  else:
    paper_count = resp[0][0]

  resp = connection.read("SELECT COUNT(id) FROM authors;")
  if len(resp) != 1 or len(resp[0]) != 1:
    author_count = 0
  else:
    author_count = resp[0][0]

  resp = connection.read("SELECT COUNT(id) FROM articles WHERE abstract IS NULL;")
  if len(resp) != 1 or len(resp[0]) != 1:
    no_abstract = 0
  else:
    no_abstract = resp[0][0]

  resp = connection.read("SELECT COUNT(id) FROM articles WHERE posted IS NULL;")
  if len(resp) != 1 or len(resp[0]) != 1:
    no_posted = 0
  else:
    no_posted = resp[0][0]

  outdated = {}
  resp = connection.read("SELECT collection, COUNT(id) FROM articles WHERE last_crawled < now() - interval %s GROUP BY collection ORDER BY collection;", (config.outdated_limit,))
  if len(resp) > 0:
    for entry in resp:
      if len(entry) < 2:
        continue # something fishy with this entry
      outdated[entry[0]] = entry[1]

  resp = connection.read("SELECT COUNT(id) FROM articles WHERE collection IS NULL;")
  if len(resp) != 1 or len(resp[0]) != 1:
    no_category = 0
  else:
    no_category = resp[0][0]

  resp = connection.read(f"""
  SELECT COUNT(id)
  FROM (
    SELECT
      a.id, COUNT(w.author) AS authors
    FROM {config.db["schema"]}.articles a
    LEFT JOIN {config.db["schema"]}.article_authors w ON w.article=a.id
    GROUP BY a.id
    ORDER BY authors
  ) AS authorcount
  WHERE authors=0;
  """)
  if len(resp) != 1 or len(resp[0]) != 1:
    no_authors = 0
  else:
    no_authors = resp[0][0]

  resp = connection.read(f"""
  SELECT COUNT(id)
  FROM (
    SELECT a.id, COUNT(z.article) AS num
    FROM prod.authors a
    LEFT JOIN prod.article_authors z ON a.id=z.author
    GROUP BY 1
    ORDER BY 2 DESC
  ) AS asdf
  WHERE num = 0
  """)
  if len(resp) != 1 or len(resp[0]) != 1:
    no_papers = 0
  else:
    no_papers = resp[0][0]

  return {
    "papers_indexed": paper_count,
    "authors_indexed": author_count,
    "missing_abstract": no_abstract,
    "missing_date": no_posted,
    "outdated_count": outdated,
    "missing_authors": no_authors,
    "missing_category": no_category,
    "authors_no_papers": no_papers
  }