112 Super ggformat
Zihan Wang
112.1 Introduction
ggformat (https://github.com/jtr13/ggformat) is an add-in tool to clean up and style ggplot2 code, and it is very useful to tidy up a single sentence when writing R code.
However, ggformat is not perfect – although it works well for a single ggplot2 sentence, it’s power is limited because it cannot handle multiple sentences, long sentences and comments. Therefore, I propose ggformat++ that can solve the aforementioned drawbacks of ggformat. The Github Repo of ggformat++ is: (https://github.com/hannawong/super_ggformat)
112.2 How it differs from ggformat
112.2.1 Handle Multiple Sentences
The most salient difference from ggformat is that ggformat++deals with multiple sentences.
The original ggformat fails to generate correct code when used on multiple sentences:
Before:
# BEFORE
library(parcoords)
library(webshot)
library(d3r)
sel_df_<-df %>%filter(Year == 2020) %>%select(County,Region,Murder,Rape,Robbery)%>%group_by(County,Region)
parcoords(data = sel_df_,brushMode = '1D-axes',color = list(colorBy = "Region"),queue = TRUE,withD3 = TRUE)There are 5 “sentences” in the code block above – the first three sentences imports libraries, and the fourth sentence modifies dataframe df, the last sentence draw a plot using parcoords. However, ggformat clearly cannot differentiate those sentences and would result in wrong answer by simply stacking all these sentences together:
##After using ggformat
library(parcoords)library(webshot)library(d3r)sel_df_<-df %>%
filter(Year == 2020) %>%
select(County,Region,Murder,Rape,Robbery)%>%
group_by(County,Region)parcoords(data = sel_df_,brushMode = '1D-axes',color = list(colorBy = "Region"),queue = TRUE,withD3 = TRUE)However, ggformat++ developed by me can identify different “sentences” and split them, ggformat++ also moves library() import sentence to the top of a code block. The output of ggformat++ is shown below, and each “sentence” is split by a new line.
## AFTER USING GGFORMAT++
library(d3r)
library(webshot)
library(parcoords)
sel_df_<-df%>%
filter(Year==2020)%>%
select(County,Region,Murder,Rape,Robbery)%>%
group_by(County,Region)
parcoords(data=sel_df_,brushMode='1D-axes',color=list(colorBy="Region"),queue=TRUE,withD3=TRUE)In the source code, I use function def prologue(full_str), def parse_sentences(str_collapse),def parse_small_sentences(large_sentence) to split sentences according to \n , <- , and right parenthesis ). Then I use function def arrange_sentence_order(atom_sentences) to rearrange library() sentence to the top of the code block.
112.2.2 Handle Comments
ggformat has trouble handling comments, for example:
##BEFORE
dt <- seattlepets %>% ###aaaaa
filter(species %in% target) %>% group_by(animal_name, species) %>% ####bbbb
summarize(n = n()) %>%mutate(s = sum(n)) %>%filter(!is.na(animal_name)) %>%ungroup()There are two comments in the code above: "###aaaaa" and "####bbbb", however, ggformat gives the wrong answer by mistakening codes as comments:
##AFTER USING GGFORMAT
dt <- seattlepets %>%
###aaaaa filter(species %in% target) %>%
group_by(animal_name, species) %>%
####bbbb summarize(n = n()) %>%
mutate(s = sum(n)) %>%
filter(!is.na(animal_name)) %>%
ungroup()I fixed this problem in ggplot++ by identifying comments:
112.2.3 Wrapping Long Sentences
In the github of ggformat, it says that ggformat has trouble wrapping long sentences.
I implement function def wrap_long_sentences(sent) in ggformat++ to split long sentence by comma, and the example mentioned above can be perfectly formatted by ggformat++:
##BEFORE
ggplot() +
geom_ribbon(data = ribbon, aes(ymin = min, ymax = max, x = x.ribbon, fill = 'lightgreen')) +
geom_line(data = ribbon, aes(x = x.ribbon, y = avg, color = 'black')) +
geom_line(data = data, aes(x = x, y = new.data, color = 'red')) +
scale_fill_identity(name = 'the fill', guide = 'legend', labels = c('m1')) +
scale_colour_manual(name = 'the colour', values = c('black' = 'black', 'red' = 'red'), labels = c('c2', 'c1')) +
xlab('x') +
ylab('density')
## AFTER USING GGFORMAT++
ggplot()+
geom_ribbon(data=ribbon,aes(ymin=min,ymax=max,x=x.ribbon,fill='lightgreen'))+
geom_line(data=ribbon,aes(x=x.ribbon,y=avg,color='black'))+
geom_line(data=data,aes(x=x,y=new.data,color='red'))+
scale_fill_identity(name='the fill',guide='legend',labels=c('m1'))+
scale_colour_manual(name='the colour',
values=c('black'='black','red'='red'),labels=c('c2','c1'))+
xlab('x')+
ylab('density')ggformat++` can also deal with extremely long sentences by wrapping it multiple times. For example, an extremely long sentence like this:
## BEFORE
ggplot() +
geom_ribbon(data = ribbon, aes(ymin = min, ymax = max, x = x.ribbon, fill = 'lightgreen')) +
geom_line(data = ribbon, aes(x = x.ribbon, y = avg, color = 'black')) +
geom_line(data = data, aes(x = x, y = new.data, color = 'red')) +
scale_fill_identity(name = 'the fill', guide = 'legend', labels = c('m1')) +
scale_colour_manual(name = 'the colour', values = c('black' = 'black', 'red' = 'red'),values = c('black' = 'black', 'red' = 'red'),values = c('black' = 'black', 'red' = 'red'),values = c('black' = 'black', 'red' = 'red'),values = c('black' = 'black', 'red' = 'red'),values = c('black' = 'black', 'red' = 'red'),
labels = c('c2', 'c1')) +
xlab('x') +
ylab('density')ggformat++ wraps the long sentence mentioned above for four times:
##AFTER USING GGFORMAT++
ggplot()+
geom_ribbon(data=ribbon,aes(ymin=min,ymax=max,x=x.ribbon,fill='lightgreen'))+
geom_line(data=ribbon,aes(x=x.ribbon,y=avg,color='black'))+
geom_line(data=data,aes(x=x,y=new.data,color='red'))+
scale_fill_identity(name='thefill',guide='legend',labels=c('m1'))+
scale_colour_manual(name='thecolour',values=c('black'='black',
'red'='red'),values=c('black'='black','red'='red'),values=c('black'='black',
'red'='red'),values=c('black'='black','red'='red'),values=c('black'='black',
'red'='red'),values=c('black'='black','red'='red'),labels=c('c2','c1'))+
xlab('x')+
ylab('density')
112.3 Ways to improve ggformat++
Only using hand-crafted rules would definitely results in some bad cases. The most reliable way to format code is to build a compiler, with which we could know the roles of each token (e.g. identifier, function, operators, comments…). Therefore, the code can format themselves according to their roles.
I have developed a compiler based on minidecaf and antlr a year ago, but it is for C++. I once thought about reusing the code to build a compiler for R, but finally give up because the workload of designing a context-free language that describe the full grammar of R is intimidating. However, I believe that the developers of R should consider adding a format module in the compiler of R, which will give much more reliable result than ggformat and ggformat++.
Moreover, I develop ggformat++ in python. In order to build it as an add-in tool of R, it needs to be transplanted into R language.