Using mutate during logging generates errors. For example, the following code generates the error:

Error in dump_log(., stop = TRUE, file = "train_log.csv") :
attempt to apply non-function

`

rm(list = ls())

train_data <- read.csv("train_original.csv",
header = TRUE,
stringsAsFactors = FALSE)
train_data <- train_data %>>%
mutate(id = seq(1, nrow(train_data), 1)) %>%
replace_with_na_all(condition = ~.x == "")

doesn't work:

train_data <- train_data %>>%
as.data.frame() %>>%
start_log(log = cellwise$new(key = "id")) %>>%
mutate(MWC = character(length = nrow(train_data))) %>>%
impute_lm(Age ~ Fare + Pclass + Sex + Embarked) %>>%
dump_log(stop = TRUE, file = "train_log.csv")

works:

train_data <- train_data %>>%
as.data.frame() %>>%
start_log(log = cellwise$new(key = "id")) %>>%
# mutate(MWC = character(length = nrow(train_data))) %>>%
impute_lm(Age ~ Fare + Pclass + Sex + Embarked) %>>%
dump_log(stop = TRUE, file = "train_log.csv")`

When an operation is performed that coerces the data into a tibble, the logging fails with errors. In the below example, the error is:

Error in $<-.data.frame(*tmp*, value, value = c("1", "2", "3", "4", :
replacement has 10692 rows, data has 12

`

library(lumberjack)
rm(list = ls())

train_data <- read.csv("train_original.csv",
header = TRUE,
stringsAsFactors = FALSE)
train_data <- train_data %>>%
mutate(id = seq(1, nrow(train_data), 1)) %>%

following ensures blanks are NA, but also coerces to tibble

replace_with_na_all(condition = ~.x == "")

doesn't work:

train_data <- train_data %>>%
#as.data.frame() %>>%
start_log(log = cellwise$new(key = "id")) %>>%
impute_lm(Age ~ Fare + Pclass + Sex + Embarked) %>>%
dump_log(stop = TRUE, file = "train_log.csv")

works

train_data <- train_data %>>%
as.data.frame() %>>%
start_log(log = cellwise$new(key = "id")) %>>%
impute_lm(Age ~ Fare + Pclass + Sex + Embarked) %>>%
dump_log(stop = TRUE, file = "train_log.csv")`

At the moment this is held back by a couple of issues: @example(s) not working and @family is ignored.

Repro with the following example, the expression is broken down to 2 columns / shift the other columns to the right and makes the logs incorrect

data(women)
women$sleutel <- 1:nrow(women)
head(women)
women <- women %L>%
  start_log(log=cellwise$new(key="sleutel")) %L>%
  mutate( istall = ifelse(height>60, "Yes", "NO")) %L>%
  dump_log(file=logfile, stop=TRUE)
head(women)

Use-case: Add username-column to log-rows to backtrack who made the changes.

Non-working solution, before dump_log():
transform( user = Sys.getenv("USERNAME")) %>>%
Results in all user-add-edits being logged, instead of username being added to existing rows in logs.

Better, but verbose solution:
After dump, read file back in, add column, write out again.

log_cellwise <- read.csv("log_cellwise.csv")
log_cellwise <- transform(log_cellwise, user = Sys.getenv("USERNAME"))
write.csv(log_cellwise, "log_cellwise.csv")

Are there some "Arguments passed to the dump method of the logger." that can be passed to achieve adding a column before the dump?
I would love if something like this was possible:
dump_log(data, "cellwise", file ="log.csv", cols = single_value/expression/transform() )
But I can see how it would open a can of worms potentially. Maybe there is some smart set of parameters / methods I am unaware of that could fit?

is it possible to use a differently named pipe instead of "%>>%" Pehaps "%>L>%"

Pipe "%>>%" is already being used by the package pipeR.

https://renkun.me/rlist-tutorial/Getting-started/Quick-overview.html

it's probably better that way..