---
title: "Introduction to logger"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Introduction to logger}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r setup, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(logger)
log_appender(appender_stdout)
```

If you are not only using R in the interactive console for ad-hoc data analysis, but running eg batch jobs (for ETL, reporting, modeling, forecasting etc) as well, then logging the status(changes) of your script is a must so that later on you can review / debug what have happened.

For most cases, it's enough to load the package and use the functions with the `log` prefix to log important and not so important messages, for example:

```{r}
library(logger)
log_info("Loading data")
data(mtcars)
log_info("The dataset includes {nrow(mtcars)} rows")
if (max(mtcars$hp) < 1000) {
  log_warn("Oh, no! There are no cars with more than 1K horsepower in the dataset :/")
  log_debug("The most powerful car is {rownames(mtcars)[which.max(mtcars$hp)]} with {max(mtcars$hp)} hp")
}
```

Interestingly, the most powerful car was not being logged -- because by default the `logger` prints messages with at least the `INFO` log level:

```{r}
log_threshold()
```

To change that, specify the new log level threshold, eg `TRACE` to log everything:

```{r}
log_threshold(TRACE)
```

The rerunning the above code chunk:

```{r}
log_info("Loading data")
data(mtcars)
log_info("The dataset includes {nrow(mtcars)} rows")
if (max(mtcars$hp) < 1000) {
  log_warn("Oh, no! There are no cars with more than 1K horsepower in the dataset :/")
  log_debug("The most powerful car is {rownames(mtcars)[which.max(mtcars$hp)]} with {max(mtcars$hp)} hp")
}
```

You may also find the `?log_eval` function useful to log both an R expression and its result in the same log record:

```{r}
f <- sqrt
g <- mean
x <- 1:31
log_eval(y <- f(g(x)), level = INFO)
str(y)
```


Sometimes, it may be reasonable to log R objects as markdown, e.g. a smallish `data.frame` or `data.table`, e.g. `mtcars` or `iris`. Calling the formatter using `pander` instead of `glue` can help:

```{r knitr-pander-setup, include = FALSE}
ppo1 <- pander::panderOptions("knitr.auto.asis")
ppo2 <- pander::panderOptions("table.style")
pander::panderOptions("knitr.auto.asis", FALSE)
pander::panderOptions("table.style", "simple")
```

```{r}
log_formatter(formatter_pander)
log_info(head(iris))
```

```{r knitr-pander-revert, include = FALSE}
pander::panderOptions("knitr.auto.asis", ppo1)
pander::panderOptions("table.style", ppo2)
```

For more details, check the [function reference in the manual](https://daroczig.github.io/logger/reference/index.html), or start with the [The Anatomy of a Log Request](https://daroczig.github.io/logger/articles/anatomy.html) and [Customizing the Format and the Destination of a Log Record](https://daroczig.github.io/logger/articles/customize_logger.html) vignettes.

```{r cleanup, include = FALSE}
logger:::namespaces_reset()
```