app.py

import pandas as pd
import streamlit as st
from collections import defaultdict
import inflect
from figures import *
from helper_fxns import *
from projections import *

# intitializes inflect class for grammar
p = inflect.engine()

# Empty Panda Dataframe that stores a read TRJ1 File
df = pd.DataFrame()

# Empty List that holds TRJ1 Objects
trj1_list = []

# Saves the count of total TRJ1 Objects
trj1_count = 0

# Set the layout of the Streamlit
st.set_page_config(page_icon=None, page_title="Counter Visualization")

# display sidebar
with st.sidebar:
    st.subheader(
        "The sidebar is resizable! Drag and drop the right border of the sidebar to resize it! ↔️")
    if file_upload := st.file_uploader(
        "Drag & drop or browse files to upload one unmodified TR-J1 spreadsheet per fiscal year:",
        type=['csv', 'tsv', 'xlsx'],
        accept_multiple_files=True,
    ):
        for file in file_upload:
            df = read_file(file, trj1_list)
        # displays files uploaded successfully using inflect module
        st.success(p.no("file", len(file_upload)) +
                            " uploaded successfully!", icon="✅")
    else:
        df = read_default_files(trj1_list)

trj1_count = len(trj1_list)
trj1_list = sort_trj1_list(trj1_list)

st.markdown("#### This app analyzes and plots TR_J1 journal usage data to allow you \
            to easily assess the usage distribution, cost per use, and usage trends \
            over time for your library's journal package subscriptions.")

with st.expander("How to use:"):
    st.write("This app takes COUNTER 5 TR_J1 reports and subscription costs for a journal package and calculates the cost per use and journal usage distribution for the time periods covered by the TR-J1 reports. TR-J1 reports contain two usage metrics: Total_Item_Requests and Unique_Item_Requests. Most librarians use the Unique_Item_Requests metric to evaluate usage, because it doesn't count repeat views and downloads of an item by the same user.")
    st.write(
        "To learn more about COUNTER 5 and TR_J1 metrics, visit: [the COUNTER website](https://www.projectcounter.org) or the [guide for librarians](https://www.projectcounter.org/wp-content/uploads/2018/03/Release5_Librarians_PDFX_20180307.pdf).")
    st.write(
        "This app was created by [Rome Duong](https://www.linkedin.com/in/phearom-d-503862195/), [Ricardo Zamora](https://www.linkedin.com/in/zamora-ricardo/), and [Clara del Junco](https://cdeljunco.github.io/me/).")

check_fiscal_year(trj1_list)
check_duplicate_dates(trj1_list)

# Accurately gets all dates for each file and saves it to a dict using dict comprehension
# key = name of file, val = list of dates
df_dates = {trj1.name: trj1.get_header_dates() for trj1 in trj1_list}

st.write("#")  # simple spacer
st.write("#")  # simple spacer

# displays count of files uploaded using inflect module
st.subheader("You have successfully uploaded " + p.no("file",
                trj1_count) + " with the following details: ")

# extract unique item requests for each TRJ1 object in trj1_list
unique_journals = [trj1.dataframe['Metric_Type'].eq('Unique_Item_Requests').sum() for trj1 in trj1_list]

# create date column for display
date_col = [f"{date_range[0].strftime('%m/%Y')} - {date_range[-1].strftime('%m/%Y')}" for date_range in df_dates.values()]

# saving file data into one hashmap
file_details = {
    "File Name": list(df_dates),
    "Date Range": list(date_col),
    "Number of Journals": list(unique_journals),
}

# creates a collapsible view of the dataframe containing in details the reporting total, the titles, and the counts of journals
with st.expander("Expand to see file details:", expanded=True):
    # convert files' data into dataframe and display it, setting to max width
    file_details_df = pd.DataFrame(file_details)
    file_details_df = file_details_df.rename_axis("Row Index")
    st.dataframe(file_details_df, use_container_width=True)

st.write("#")  # simple spacer
# display tabs of dataframes in expander
with st.expander("Expand to see raw TR_J1 data:"):
    if not date_col:
        st.write("Please upload data")
    else:
        tabs = st.tabs(date_col)
        for i, trj1 in enumerate(trj1_list):
            with tabs[i]:
                st.dataframe(trj1.dataframe)

st.write("#")  # simple spacer

# Decision Tree to verify that both metric types exist in trj1 files
if 'Unique_Item_Requests' in df['Metric_Type'].values and "Total_Item_Requests" in df['Metric_Type'].values:
    metric_choice = st.radio(
        "Select which metric type to use:",
        ("Unique Item Requests", "Total Item Requests")
    )
else:
    st.error(st.warning(
        'Please make sure that you have a valid metric type of Total Item Requests or Unique Item Requests', 
                icon="⚠️"))
# only include rows for specified metric choice
trj1_list = update_metric_choice(trj1_list, metric_choice)

# Filter incomplete/complete TRJ1's
complete_trj1_list, incomplete_trj1 = set_complete_incomplete_files(trj1_list)

# Create projections for each incomplete TRJ1 
for incomplete in incomplete_trj1:
    projection = project_total_uses(complete_trj1_list, incomplete)
    incomplete.set_projected_usage(projection)

# sidebar (Cost Per Use: Input and Output)
st.sidebar.write("#")  # simple spacer
st.sidebar.header("Cost Per Use")
st.sidebar.write('Input the journal package cost in dollars for the period covered by each TR_J1 file:')
cpu_list = []

for i, trj1 in enumerate(trj1_list):
    cost = st.sidebar.number_input(
        ' ' + date_col[i] + ' ', min_value=0.00, format="%f", key=trj1.name)
    if cost != 0 and trj1.is_Full_FY():
        trj1.set_cost_per_use(cost) 
        cpu_list.append(trj1.cpu)
    elif cost != 0 and not trj1.is_Full_FY():
        trj1.set_projected_cost_per_use(cost)
        cpu_list.append(trj1.projected_cpu)
    else: 
        trj1.cpu = 0
        trj1.projected_cpu = 0
        cpu_list = [0 for _ in trj1_list]

    st.sidebar.write("The Cost Per Use is : $ " + format(trj1.cpu if trj1.is_Full_FY() else trj1.projected_cpu, ".2f") + ("*" if not trj1.is_Full_FY() else ""))
if incomplete_trj1:
    st.sidebar.write("*Data provided for this year is incomplete. \
                            This is the projected CPU for the entire fiscal year \
                            based on patterns of use in the other years of data provided.")

############### Streamlit: Displaying Data #################
# using counter to get occurences of each num
titles_set = set()
max_df_values = []

title_header = "Journals with this many item requests(double-click to see full list)"
# loop through each dataframe in the list to create new dataframe
metric_choice = "Reporting_Period_Total"
count_header = "Number of journals with this many item requests"
title_header = "Titles of journals with this many item requests"

# Define a defaultdict to store the titles of the journals
titles = defaultdict(list)
titles_set = set()

# Iterate through the list of trj1 objects
occurrences_list = []

for trj1 in trj1_list:

    df = trj1.dataframe
    # Count the number of occurrences of each value in the metric_choice column
    occurrences = df[metric_choice].value_counts().sort_index()

    # Group the data by the values in the metric_choice column
    grouped_data = df.groupby(metric_choice)

    # Apply a lambda function to each group to create a list of titles
    titles_list = grouped_data.apply(lambda x: x['Title'].tolist())

    # Combine the occurrences and titles_list into a single DataFrame
    usage_df = pd.concat([occurrences, titles_list], axis=1).reset_index()

    # Rename columns
    usage_df.columns = [metric_choice, count_header, title_header]

    # Store the titles in a defaultdict
    for index, row in usage_df.iterrows():
        titles[row[metric_choice]].extend(row[title_header])
        titles_set.update(row[title_header])
    occurrences_list.append(usage_df)

# Create Line plot of Distribution of Cost Per Use
st.header("Distribution of Cost Per Use")
if date_col and 0 not in cpu_list:
    fig1 = linePlot(date_col, cpu_list)[0]
    fig1.update_layout(yaxis=dict(range=[0, max(linePlot(date_col, cpu_list)[1]["Cost Per Use"])+1]))
    st.plotly_chart(fig1)
else:
    st.warning("Please provide input for Cost Per Use in the sidebar")


# Usage Distribution - Tabs
st.write("#")  # simple spacer
st.header("Usage Distribution")
st.write("See which journals were used the most and least for each \
            of the time periods covered by your TR_J1 reports")

if not date_col:
    st.write("Please provide a data input")
else:
    hist_tab = st.tabs(date_col)
    # precompute max_report across all dataframes
    max_reports = [int(occurrences[metric_choice].max()) for occurrences in occurrences_list]
    for i, trj1 in enumerate(trj1_list):
        with hist_tab[i]:
            # get the dataframe from the list created earlier so it can be displayed
            usage_df = occurrences_list[i].rename_axis("Row Index")
            max_count = usage_df[count_header].max()

            # create a filter slider and use user input to create a filtered dataframe
            filter_slider = st.slider("Set the minimum and maximum reporting period total (x-axis) here.", 1, max_reports[i], value=(1, max_reports[i]), key=i)
            filter_min = filter_slider[0]
            filter_max = filter_slider[1]
            filtered_df = trj1.dataframe.query('Reporting_Period_Total >= @filter_min and Reporting_Period_Total <= @filter_max')

            # display number of filtered journals
            filter_diff = filter_max - filter_min
            filter_count = len(filtered_df)
            
            # grammar based on filter slider using inflect
            if filter_diff:
                st.write("There " + p.plural("is", filter_count) + " currently " + p.no("journal", filter_count) +
                        " within the following range: {} - {} reporting period total.".format(filter_min, filter_max))
            else:
                st.write("There " + p.plural("is", filter_count) + " currently " + p.no(
                    "journal", filter_count) + " with a {} reporting period total.".format(filter_min))


            # determine the height for the histogram based on the maximum count
            chart_height = 600 if 300 <= max_count <= 600 else max_count if max_count < 300 else 500

            # create a collapsible view of the filtered dataframe
            with st.expander("Expand to see the filtered data behind the distribution:", expanded=False):
                st.write("#")
                st.caption("Click on column header to sort by ascending/descending order")
                st.dataframe(filtered_df, use_container_width=True)

            # display the histogram
            stacked_hist = histogram(filtered_df, filter_slider[1], chart_height)
            st.write("#")
            st.altair_chart(stacked_hist, use_container_width=True)

st.write("Click on plot and scroll to zoom, click & drag to move, \
            and double-click to reset view. Click ... at top right \
            to download the chart as an SVG/PNG.")

# Create a bar chart showing the reporting period for specific journals
st.header("Reporting Period Total over Time")
# multiselect option for titles
titles_selected = st.multiselect(
    "Search and click on the titles you want to view in the bar chart.",
    sorted(titles_set)
)  # contains a max selections param if we want the user to only select a limited amount

bar_df = []
bar_df.extend(titles_selected)

# determine whether the user has selected any journal
if not titles_selected:
    st.warning("To view journals over time, please select journals above.")
else:
    # create a new dataframe that combines all dataframes together and add a column of Fiscal Year to the dataframe corresponding to their fiscal year
    concat_df = pd.concat([df.dataframe.assign(Fiscal_Year=date_col[i]) for i, df in enumerate(trj1_list)], ignore_index=True)
    if bar_df:
        # filter the dataframe by the selected titles
        df = concat_df[concat_df["Title"].isin(bar_df)]
        # convert the fiscal year column into a categorical variable and create a Plotly bar chart
        df['Fiscal_Year'] = pd.Categorical(df['Fiscal_Year'])
        fig = barChart(df)
        st.plotly_chart(fig)
    else:
        st.warning("Please select at least one title to view journals over time.")


# st.write([trj1.rpt for trj1 in trj1_list])
# remaining_months = calculate_remaining_months(incomplete_trj1)
# st.write(remaining_months)