TITLE: Analysing Ledger Personal Accounting Data Using R
DATE: 2017-09-01
AUTHOR: John L. Godlee
====================================================================


[Ledger] (aka Ledger-CLI) provides a great framework to keep track
of your personal finances. Keeping your data in plain text format
provides multiple benefits:

  [Ledger]: http://ledger-cli.org

-   You can read the file without any fancy programs, just a text
    editor
-   Even if Ledger stops being maintained, you can still use the
    journal file
-   The file can be interpreted using many other programming
    languages, like R!

To properly visualise how my finances are changing over time
however, I find the text based reports provided by ledger-cli a bit
dense and fiddly.

This post might be an exercise in reinventing the wheel. Ledger
already has a decent [web-based reporting system] that can provide
pretty graphs and lots of other ledger-like apps that can do
similar. But my language of choice for making pretty graphs and
manipulating data is ‘R’, so I’m going to use that.

  [web-based reporting system]: https://github.com/ledger/ledger-web

ost of the code below is actually data manipulation, which I’ve
chosen to do with the dplyr package, creating the plots in ggplot2
isn’t too taxing. I’ve created an example script [here] and you
can find the example .ledger.journal file I used [here][1]

  [here]: https://johngodlee.github.io/files/ledger/ledger_journal_analysis.R
  [1]: https://johngodlee.github.io/files/ledger/example_ledger.journal

Firstly, export your ledger journal file (.ledger.journal) as a .csv
in the terminal, the name and filepath of your journal file might be
different:

    touch ledger.csv

    ledger csv -f ~/.ledger.journal > ledger.csv

And that’s the last thing we’ll be doing in the shell, everything
else will be in R. So fire up an R session to start analysing the
data.

Firstly set the working directory, import the csv file and load some
packages:

    # set working directory to `.ledger.journal`
    setwd()

    # Create vector of column names
    journal_names <- c("Date", "NA_1", "Description", "Source", "Currency", "Amount", "NA_2", "NA_3")

    # Import csv
    ledger <- read.csv("ledger.csv", col.names = journal_names)

    # Load packages
    library(dplyr)
    library(ggplot2)

Now to make the ledger dataframe easier to use:

    # Convert "Date" column to date class
    ledger$Good_date <- as.Date(ledger$Date, format = "%Y/%m/%d")
    class(ledger$Good_date)  # To check the above worked

    # Sort by "Good_date"
    ledger_sort <- ledger[order(ledger$Good_date),]

    # Add cumulative column for each source
    ledger_cumsum <- ledger_sort %>%
        group_by(Source) %>%
        mutate(Cumulative = cumsum(Amount))

The rest involves creating a few graphs that I find useful. For all
these plots to work in their current form however, your Source or
“Account” structure must be the similar to the recommended structure
found in the [ledger-cli example journal], e.g.:

  [ledger-cli example journal]: http://ledger-cli.org/3.0/doc/ledger3.html#Example-Journal-File

    ┃
    ┣Assets
    ┃┣Checking
    ┃┣Savings
    ┃┗Cash
    ┣Income
    ┃┣Work
    ┃┗Ebay_sales
    ┗Expenses
     ┣Socialising
     ┣Bills
     ┗Mortgage

For instance if I had to pay a bill in ledger, the journal entry
might look like this:

    2017/12/06 Electricity bill
        Assets:Checking         $-65.51
        Expenses:Bills          $ 65.51

But it should be trivial to change the code to match your journal
structure.

Assets over time

  {IMAGE}


Create a data frame only containing assets:

    assets <- ledger_cumsum %>%
        filter(grepl("Assets", Source))

Then make the plot:

    ggplot(assets, aes(x = Good_date, y = Cumulative, group = Source)) +
        geom_hline(aes(yintercept = 0), colour = "red") +
        geom_line(aes(colour = Source), size = 1.2) +
        geom_point(aes(colour = Source), size = 2) +
        scale_x_date(date_breaks = "1 week", date_labels = "%W/%y")

Viewing a particular asset in detail over time

  {IMAGE}


    # Create data frame
    assets_bank_current <- ledger_cumsum %>%
        filter(Source == "Assets:Checking")

    # Line plot of student account over time with description of expenditure
    ggplot(assets_bank_current, aes(x = Good_date, y = Cumulative, group = Source, label = Description)) + 
        geom_line() + 
        geom_text() + 
        scale_x_date(date_breaks = "2 days", date_labels = "%W/%y") + 
        xlab("Date WW-YY") + 
        ylab("Balance ($)")

Bar plots with breakdown of expenses

  {IMAGE}


  {IMAGE}


    # Create summary dataframe of expenses
    expenses_sum <- ledger_cumsum %>%
        filter(grepl("Expenses", Source)) %>%
        group_by(Source) %>%
        summarise(Amount = sum(Amount)) %>%
        mutate(Percentage = Amount / sum(Amount) * 100) %>%
        mutate(Source = factor(Source, levels = Source[order(Amount, decreasing = TRUE)]))  # Create ordered factor for x axis

    # Bar plot
    ggplot(expenses_sum, aes(x = Source, y = Amount)) + 
        geom_bar(stat = "identity", aes(fill = Source)) + 
        theme(legend.position = "none") +
        ylab("Amount ($)")

    # Stacked percentage bar chart
    ggplot(expenses_sum, aes(x = NA, y = Percentage, fill = Source)) + 
        geom_bar(stat = "identity") + 
        geom_text(aes(label = paste(round(Percentage, digits = 2), "% - ", Source, sep="")), position=position_stack(vjust=0.5))

Last 30 days income/expenses summary

  {IMAGE}


Creating this plot was fun, I had a go at using ifelse() arguments
inside the ggplot() call in order to change the position of an error
bar and text (which I’ve used to show deficit) depending on whether
I’ve made a net gain or loss that month.

    # Create summary dataframe
    ledger_30d_summ <- ledger_cumsum %>%
        filter(Good_date > as.Date(Sys.Date(), format = "%Y-%m-%d") - 30) %>%
        filter(grepl("Assets", Source)) %>%
        mutate(expense_income = if_else(Amount > 0, "Income", "Expense")) %>%
        group_by(expense_income) %>%
        summarise(Total = sum(Amount)) %>%
        mutate(Total = abs(Total))


    # Create colour palette 
    expense_income_palette <- c("#D43131", "#1CB5DB")

    # Create plot
    ggplot(ledger_30d_summ, aes(x = expense_income, y = Total), environment = environment()) + 
        geom_bar(stat = "identity", fill = expense_income_palette) + 
        geom_errorbar(aes(x = ifelse(ledger_30d_summ$Total[1] > ledger_30d_summ$Total[2], "Income", "Expense"), 
                ymax = max(ledger_30d_summ$Total),
                ymin = min(ledger_30d_summ$Total))) +
        geom_text(aes(x = ifelse(ledger_30d_summ$Total[1] > ledger_30d_summ$Total[2], "Income", "Expense"),
            y = min(ledger_30d_summ$Total) + 0.5*(max(ledger_30d_summ$Total) - min(ledger_30d_summ$Total)),
            label = ifelse(ledger_30d_summ$Total[1] > ledger_30d_summ$Total[2],
            paste("$ -", max(ledger_30d_summ$Total) - min(ledger_30d_summ$Total), sep = ""),
            paste("$ ", max(ledger_30d_summ$Total) - min(ledger_30d_summ$Total), sep = "")),
            hjust = -0.5)) + 
        xlab("Expense/Income") + 
        ylab("Amount ($)") 

Now that I’ve defined all these plots, it shouldn’t take too much
effort to turn them into a basic Shiny app that I can load up in my
web browser, or run a script that saves the plots as images on my
computer so I can look at them later.