arrange.Rmd

---
title: ""
author: ""
date: ""
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE, df_print="info_paged")
```

```{r, echo=FALSE}
info_paged_print <- function(x, options) {
  tibble_info <- paste0("<div class=\"tibble-info\">A tibble: ", nrow(x), " x ", ncol(x), "</div>")
  group_info <- paste0("<div class=\"group-info\">Groups: ", 
                      paste0(group_vars(x), collapse = ", "), 
                      " [", nrow(group_keys(x)), "]", "</div>")
  
  if (dplyr::is_grouped_df(x)) {
    tab_info <- paste0("<div class=\"info\">", tibble_info, " ", group_info, "</div>")
    cat(tab_info)
  } else {
    cat(paste0("<div class=\"info\">", tibble_info, "</div>"))
  }
  knitr::asis_output(
    rmarkdown:::paged_table_html(x, options = attr(x, "options")),
    meta = list(
      dependencies = rmarkdown:::html_dependency_pagedtable()
    )
  )
}

knitr::opts_hooks$set(df_print = function(options) {
  if (options$df_print == "info_paged") {
    options$render = info_paged_print
    options$comment = ""
    options$results = "asis"
  }
  options
})
```

```{css, echo=FALSE}
.tibble-info,
.group-info {
  display: inline-block;
  padding: 15px;
}

.info {
  margin-top: 5px;
  margin-bottom: 5px;
  border: 1px solid #ccc;
  border-radius: 4px;
  font-weight: 600;
  color: #999898;
}
```

```{r, message = FALSE, echo=FALSE}
library(dplyr)
library(readxl)
df <- read_excel(here::here("online_retail_II.xlsx"))
```

data-masking

# - *fundamentals*

`arrange()` reorders the rows of a data frame by the values of one or several columns. 

In case of numerical ones, the default puts the smallest values first.

```{r}
df %>%
  arrange(Quantity)
```

We need to use a minus (`-`) or to wrap the column in `desc()` if we are interested in seeing the largest ones on top.

```{r}
df %>%
  arrange(-Quantity)
df %>%
  arrange(desc(Quantity))
```

The documentation doesn't specify what happens in case of ties but we can assume that rows with a smaller row index come first, preserving their original row order then.

In case of character columns, the order is alphabetical as defined by the default "C" locale.

```{r}
df %>%
  arrange(Description)
df %>%
  arrange(desc(Description))
```

# - *.locale*

We can change it to the locale of our choice with the `.locale` argument.

```{r}
df %>%
  arrange(Description, .locale = "en")
```

In case we want to arrange character columns by a custom order, it is possible by transforming them in ordered factors.

```{r}
df %>%
  mutate(Description_Factor = factor(Description, levels = sample(unique(df$Description), 
                                                                  length(unique(df$Description))), ordered = TRUE)) %>%
  arrange(Description_Factor)
```

`arrange()` is a data-masking function, so it accepts expressions as input.

```{r}
df %>%
  arrange(Quantity ^ 2)
df %>%
  arrange(min_rank(Quantity))
```

NAs are placed at the bottom, but we can have them on top by using this syntax (that exploits the data-masking nature of `arrange()` putting the FALSEs, that are equal to 0, first).

```{r}
df %>%
  arrange(!is.na(Description))
```

If we use more columns, the subsequent ones will be used to resolve ties (notice how rows 2 and 3 swap place in the second line of code).

```{r}
df %>%
  arrange(desc(Quantity))
df %>%
  arrange(desc(Quantity), StockCode)
```

# - *with group_by()*

`arrange()` ignores grouping. In this way we can have rows from different groups at the top.

```{r}
df %>%
  group_by(Description) %>%
  arrange(Quantity)
```

In case we want to order by the grouping column as well, we need to add it, before or after the other columns as the situation requires.

```{r}
df %>%
  group_by(Description) %>%
  arrange(Quantity, Description)
df %>%
  group_by(Description) %>%
  arrange(Description, Quantity)
```

## - *.by_group*

Instead of specifying the grouping column, we can set the `.by_group` argument to TRUE.
This will use the grouping column as the first one specified, like in the previous example.

```{r}
df %>%
  group_by(Description) %>% 
  arrange(Quantity, .by_group = TRUE)
```

## - *using expressions*

If we use an expression within `arrange()`, that will be evaluated on all the data frame and non per groups.  
Therefore in the following example the verb doesn't modify the rows' order as `sum(Quantity)` returns the same value for all the rows.

```{r}
df %>%
  group_by(Description) %>%
  arrange(desc(sum(Quantity)))
```

So we need to be explicit with a `mutate()` call and use the resulting additional column to arrange by.

```{r}
df %>%
  group_by(Description) %>%
  mutate(Total_Quantity = sum(Quantity)) %>%
  arrange(desc(Total_Quantity))
```