Harvest information about newspaper issues

When you search Trove's newspapers, you find articles – these articles are grouped by page, and all the pages from a particular date make up an issue. But how do you find out what issues are available? On what dates were newspapers published? This notebook shows how you can get information about issues from the Trove API.

The code below generates two datasets:

  • Total number of issues per year for every newspaper – 27,615 rows with the fields:
    • title – newspaper title
    • title_id – newspaper id
    • state – place of publication
    • year – year published
    • issues – number of issues
  • Complete list of issues for every newspaper – 2,655,664 rows with the fields:
    • title – newspaper title
    • title_id – newspaper id
    • state – place of publication
    • issue_id – issue identifier
    • issue_date – date of publication (YYYY-MM-DD)

These were harvested on 18 October 2021. You can download the pre-harvested datasets from CloudStor:

Issue urls

To keep the file size down, I haven't included an issue_url in the issues dataset, but these are easily generated from the issue_id. Just add the issue_id to the end of http://nla.gov.au/nla.news-issue. For example: http://nla.gov.au/nla.news-issue495426. Note that when you follow an issue url, you actually get redirected to the url of the first page in the issue.

In [1]:
import requests
from requests.adapters import HTTPAdapter
from requests.packages.urllib3.util.retry import Retry
import arrow
import requests_cache
from tqdm.auto import tqdm
import time
import pandas as pd
import json
import altair as alt

# Create a session that will automatically retry on server errors
s = requests_cache.CachedSession()
retries = Retry(total=5, backoff_factor=1, status_forcelist=[ 502, 503, 504 ])
s.mount('http://', HTTPAdapter(max_retries=retries))
s.mount('https://', HTTPAdapter(max_retries=retries))
In [20]:
API_KEY = 'YOUR API KEY'
API_URL = 'https://api.trove.nla.gov.au/v2/newspaper/title/'

PARAMS = {
        'encoding': 'json',
        'key': API_KEY
}

Total number of issues per year for every newspaper in Trove

To get a list of all the newspapers in Trove you make a request to the newspaper/titles endpoint. This provides summary information about each title, but no data about issues.

To get issue data you have to request information about each title separately, using the newspaper/title/[title id] endpoint. If you add include=years to the request, you get a list of years in which issues were published, and a total number of issues for each year. We can use this to aggregate information about the number of issues by title and year.

In [11]:
def get_issues_by_year():
    '''
    Gets the total number of issues per year for each newspaper.
    
    Returns:
      * A list of dicts, each containing the number of issues available from a newspaper in a particular year
    '''
    years = []
    
    # First we get a list of all the newspapers (and gazettes) in Trove
    response = s.get('https://api.trove.nla.gov.au/v2/newspaper/titles/', params=PARAMS)
    data = response.json()
    titles = data['response']['records']['newspaper']
    
    # Then we loop through all the newspapers to retrieve issue data
    for title in tqdm(titles):
        params = PARAMS.copy()
        
        # This parameter adds the number of issues per year to the newspaper data
        params['include'] = 'years'
        response = s.get(f'{API_URL}{title["id"]}', params=params)
        try:
            data = response.json()
        except json.JSONDecodeError:
            print(response.url)
            print(response.text)
        else:
            # Loop through all the years, saving the totals
            for year in data['newspaper']['year']:
                years.append({
                    'title': title['title'],
                    'title_id': title['id'],
                    'state': title['state'],
                    'year': year['date'],
                    'issues': int(year['issuecount'])
                })
        if not response.from_cache:
            time.sleep(0.2)
    return years
In [ ]:
issue_totals = get_issues_by_year()
In [13]:
# Save results as a dataframe
df_totals = pd.DataFrame(issue_totals)
df_totals.head()
Out[13]:
title title_id state year issues
0 Canberra Community News (ACT : 1925 - 1927) 166 ACT 1925 3
1 Canberra Community News (ACT : 1925 - 1927) 166 ACT 1926 12
2 Canberra Community News (ACT : 1925 - 1927) 166 ACT 1927 9
3 Canberra Illustrated: A Quarterly Magazine (AC... 165 ACT 1925 1
4 Federal Capital Pioneer (Canberra, ACT : 1924 ... 69 ACT 1924 1

How many issues are there?

In [14]:
df_totals['issues'].sum()
Out[14]:
2655664
In [15]:
df_totals.shape
Out[15]:
(27615, 5)
In [16]:
# Save as a CSV file
df_totals.to_csv(f'newspaper_issues_totals_by_year_{arrow.now().format("YYYYMMDD")}.csv', index=False)

Display the total number of issues per year

By grouping the number of issues by year, we can see how the number of issues in Trove changes over time. It's interesting to compare this to the number of articles over time.

In [17]:
# Group by year and calculate sum of totals
df_years = df_totals.groupby(by='year').sum().reset_index()
In [18]:
# Create a chart
alt.Chart(df_years).mark_bar().encode(
    x='year:O',
    y='issues:Q',
    tooltip=['year:O', 'issues:Q']
).properties(width=800)
Out[18]:

Harvest a complete list of issues

We've found out how many issues were published, but not when they were published. To get a complete list of issue dates and identifiers we have to add another parameter to our title API request. The range parameter sets a date range. If we add it to our request, the API will return information about all the issues within that date range.

How do we set the range? The summary inforation for each title includes startDate and endDate fields. We could simply set the range using these, however, this could return a huge amount of data. It's best to be conservative, requesting the issue data in manageable chunks. The code below iterates over the complete date range for each title, requesting a year's worth of issues at a time. Note that the range parameter expects a date range in the format YYYYMMDD-YYYYMMDD.

It turned out that some titles don't have start and end dates, and some of the start and end dates are wrong. I've found ways to work around these. See below for more information.

In [5]:
# These are newspapers where the date ranges are off by more than a year
# In these cases we'll harvest all the issues in one hit, rather than year by year
dodgy_dates = ['1486', '1618', '586']

def get_title_summary(title_id):
    '''
    Get the details of a single newspaper title.
    '''
    response = s.get(f'{API_URL}{title_id}', params=PARAMS)
    data = response.json()
    return data['newspaper']

def get_issues_in_range(title_id, start_date, end_date):
    '''
    Get a list of issues available from a particular newspaper within the given date range.
    '''
    issues = []
    params = PARAMS.copy()
    params['include'] = 'years'
    params['range'] = f'{start_date.format("YYYYMMDD")}-{end_date.format("YYYYMMDD")}'
    response = s.get(f'{API_URL}{title_id}', params=params)
    try:
        data = response.json()
    except json.JSONDecodeError:
        print(response.url)
        print(response.text)
    else:
        for year in data['newspaper']['year']:
            if 'issue' in year:
                for issue in year['issue']:
                    issues.append({
                        'title_id': title_id,
                        'issue_id': issue['id'],
                        'issue_date': issue['date']
                    })
    if not response.from_cache:
        time.sleep(0.2)
    return issues

def get_issues_full_range(title_id):
    '''
    In most cases we set date ranges to get issue data in friendly chunks. But sometimes the date ranges are missing or wrong.
    In these cases, we ask for everything at once, by setting the range to the limits of Trove.
    '''
    start_date = arrow.get('1803-01-01')
    range_end = arrow.now()
    issues = get_issues_in_range(title_id, start_date, range_end)
    return issues

def get_issues_from_title(title_id):
    '''
    Get a list of all the issues available for a particular newspaper.
    
    Params:
      * title_id - a newspaper identifier
    Returns:
      * A list containing details of available issues
    '''
    issues = []
    title_summary = get_title_summary(title_id)
    
    # Date range is off by more than a year, so get everything in one hit
    if title_id in dodgy_dates:
        issues += get_issues_full_range(title_id)
    else:
        try:
            # The date ranges are not always reliable, so to make sure we get everything
            # we'll set the range to the beginning and end of the given year
            start_date = arrow.get(title_summary['startDate']).replace(day=1, month=1)
            end_date = arrow.get(title_summary['endDate']).replace(day=31, month=12)
        except KeyError:
            # Some records have no start and end dates at all
            # In this case set the range to the full range of Trove's newspapers
            issues += get_issues_full_range(title_id)
        else:
            # If the date range is available, loop through it by year
            while start_date <= end_date:
                range_end = start_date.replace(month=12, day=31)
                issues += get_issues_in_range(title_id, start_date, range_end)
                start_date = start_date.shift(years=+1).replace(month=1, day=1)
    return issues

def get_all_issues():
    issues = []
    response = s.get('https://api.trove.nla.gov.au/v2/newspaper/titles/', params=PARAMS)
    data = response.json()
    titles = data['response']['records']['newspaper']
    for title in tqdm(titles):
        title_issues = get_issues_from_title(title['id'])
        issues += [dict(i, title=title['title'], state=title['state']) for i in title_issues]
    return issues
In [ ]:
issues = get_all_issues()
In [7]:
len(issues)
Out[7]:
2655664
In [8]:
df_issues = pd.DataFrame(issues)
df_issues.head()
Out[8]:
title_id issue_id issue_date title state
0 166 495445 1925-10-14 Canberra Community News (ACT : 1925 - 1927) ACT
1 166 495422 1925-11-11 Canberra Community News (ACT : 1925 - 1927) ACT
2 166 495423 1925-12-11 Canberra Community News (ACT : 1925 - 1927) ACT
3 166 495424 1926-01-11 Canberra Community News (ACT : 1925 - 1927) ACT
4 166 495425 1926-02-11 Canberra Community News (ACT : 1925 - 1927) ACT
In [9]:
df_issues.to_csv(f'newspaper_issues_{arrow.now().format("YYYYMMDD")}.csv', index=False)

Check to see what's missing

I ran the code below a few times in order to identify problems with the harvest. It helped me track down newspapers that had errors in the date ranges. Most of the errors were small and I could pick up any missing issues by expanding the date range to cover a whole year. But in some cases, the date ranges missed multiple years. To get the issues in these missing years, I created a dodgy_dates list. Any newspapers in this list are processed differently – the given date range is ignored, and instead the range is set to cover the period from 1803 to the present! Titles that are missing start and end dates are treated the same way. Once these fixes were included, the only missing issues left were from the Noosa News – requesting issues from this newspaper originally caused an error, but this bug has now been fixed by Trove so no issues are missing!

In [19]:
# Compare the total number of issues reported by the API with the number actually harvested
# This helps us identify cases where the harvest has failed for some reason.
missing = 0
for title, years in df_totals.groupby(by=['title_id', 'title']):
    num_issues = df_issues.loc[df_issues['title_id'] == title[0]].shape[0]
    if years['issues'].sum() != num_issues:
        print(title[0], title[1])
        print(f'Year totals: {years["issues"].sum()}')
        print(f'Issues harvested: {num_issues}')
        missing += (years["issues"].sum() - num_issues)

Created by Tim Sherratt for the GLAM Workbench. Support this project by becoming a GitHub sponsor.