Ryan Price ryapric@gmail.com
loggit is an ndJSON logging
library for R software. It is blazingly fast when writing logs, and has
zero external dependencies. loggit can be as simple and unobstrusive
as you’d like, or as involved as your application needs it to be.
Please see below for some quick examples, and read the vignettes for the Getting Started guide, as well as some other use case examples.
There are indeed several logging packages available for R. loggit,
however, takes a more modern approach approach to logging in R:
- Opting to use the JSON format, which is parsable by most modern software
- Designed with log streams in mind
- Unobtrusive, yet highly flexible
- Convenient ability to log data, then analyze that log data on the same host.
Additionally, the boilerplate to get going with loggit is minimal at
worst, only requiring you to point to the log file. If deploying your R
code in a container ecosystem, you don’t even need to do that, since
loggit will echo its formatted logs to stdout. No need to write
custom formatters, handlers, levels, etc. – just f&ck#n’
loggit!
The quickest way to get up & running with loggit is to… do nothing
special. loggit’s simplest functionality does its best to stay out of
your way. You’ll probably want to point it to a log file, though;
otherwise, logs will print to the console, but land in a tempfile.
library(loggit)
#>
#> Attaching package: 'loggit'
#> The following objects are masked from 'package:base':
#>
#> message, stop, warning
# set_logfile("./loggit.log")
message("This is a message")
#> {"timestamp": "2022-04-12T10:55:02-0500", "log_lvl": "INFO", "log_msg": "This is a message"}
#> This is a message
warning("This is a warning")
#> {"timestamp": "2022-04-12T10:55:02-0500", "log_lvl": "WARN", "log_msg": "This is a warning"}
#> Warning in warning("This is a warning"): This is a warning
# stop("This actually throws a critical error, so I'm not actually going to run it here :)"))
#> {"timestamp": "2020-05-31T20:59:33-0500", "log_lvl": "ERROR", "log_msg": "This actually throws a critical error, so I'm not actually going to run it here :)"}You can suppress each part of the console output separately (both the
loggit ndJSON and the regular R output) but the default is to post
both. Only the ndJSON is written to the log file.
You can also use the loggit() function directly to compose much more
custom logs, including entirely custom fields (and prevent
throwing actual status codes until you wish to handle them). loggit
doesn’t require that you set custom logger objects or anything like
that: just throw whatever you want at it, and it’ll become a structured
log.
loggit("ERROR", "This will log an error - but not actually throw one yet", rows = nrow(iris), anything_else = "you want to include")
#> {"timestamp": "2022-04-12T10:55:02-0500", "log_lvl": "ERROR", "log_msg": "This will log an error - but not actually throw one yet", "rows": "150", "anything_else": "you want to include"}
# Read log file into data frame to implement logic based on entries
logdata <- read_logs()
print(logdata)
#> timestamp log_lvl
#> 1 2022-04-12T10:55:02-0500 INFO
#> 2 2022-04-12T10:55:02-0500 WARN
#> 3 2022-04-12T10:55:02-0500 ERROR
#> log_msg rows
#> 1 This is a message
#> 2 This is a warning
#> 3 This will log an error - but not actually throw one yet 150
#> anything_else
#> 1
#> 2
#> 3 you want to include
if (any(logdata$log_lvl == "ERROR")) {
print("Errors detected somewhere in your code!") # but you can throw a stop() here, too, for example
}
#> [1] "Errors detected somewhere in your code!"Again, check out the vignettes for more details!
You can install the latest CRAN release of loggit via
install.packages("loggit").
Or, to get the latest development version from GitHub –
Via devtools:
devtools::install_github("ryapric/loggit")
Or, clone & build from source:
cd /path/to/your/repos
git clone https://github.com/ryapric/loggit.git loggit
make install
To use the most recent development version of loggit in your own
package, you can include it in your Remotes: field in your DESCRIPTION
file:
Remotes: github::ryapric/loggit
Note that packages being submitted to CRAN cannot have a Remotes
field. Refer
here
for more info.
MIT