R for python people¶
Summary¶
When I was just started my Ph.D, I was using Perl, then I moved to python and started to learn some R as well as other languages such as Matlab. Then I decided to use just one programming language (python) because it just feels too inconvenient to remember and use the same thing but in many different ways. And now, as job needed, I have to pick up some R.
Just like how I moved from Perl to Python, the key is to get familiar with the syntax for common logic and packages. In this post, I will gradually list them here:
A unique command in R is %>%, which is similar to the pipe function in Linux |.
R plots¶
I’m not a good R user. Some of the following plots took me hours to figure out.
In this barplot, I have made clusters for WT and KO cells, I then want to make a barplot showing the “normalized value”, basically cluster size divided by the corresponding WT or KO size. The key here is to use aes( y=..count../..Total_cell..) and a custom column in the aes mapping function.
a=cell_data %>%
group_by(orig.ident) %>%
summarise(Total_cell = n())
cell_data2 = merge(cell_data,a,by=c("orig.ident"))
ggplot(cell_data2, aes(x = manual_label, fill = orig.ident,Total_cell=Total_cell)) +
geom_bar(aes( y=..count../..Total_cell..),position=position_dodge2(reverse = TRUE),size=0.6,alpha=0.5,color="black") +
scale_x_discrete(limits =new_level)+
scale_fill_manual(values = c("#0073C2FF","#EFC000FF"))+theme(axis.text.x = element_text(angle = 45, vjust =1, hjust=1),
text = element_text(size=12))
simple command differences¶
length(x) len(x)
glue("xxx{asd}") f"xxx{asd}"
logic flow differences¶
for loop R
n=10
s=c(1,2,3)
for (i in 1:n){
current_s = s[i]
}
common package differences¶
Ref:
https://shiring.github.io/r_vs_python/2017/01/22/R_vs_Py_post
data frame operation¶
Replace a column value using values from another column¶
https://stackoverflow.com/questions/40177132/replace-values-from-another-dataframe-by-ids
library("dplyr")
as1 <- data.frame(ID = c(1,2,3,4,5,6), pid = c(21,22,23,24,25,26),Values = c(435,33,45,NA, NA,12))
as2 <- data.frame(ID = c(4,5),pid = c(24,25), Values = c(544, 676))
left_join(as1, as2, by = c("ID", "pid")) %>%
mutate(Values = ifelse(is.na(Values.x), Values.y, Values.x)) %>%
select(ID, pid, Values)
For my data, the left_join command produced NA values in the new columns, I have to do this:
label = read.csv("info.tsv",sep="\t",header=F,stringsAsFactors=F)
Python ↔R basics # Python ⇔ R: object types type(a) ⇔ typeof(a) # Python ⇔ R: variable assignment a=5 ⇔ a<-5 # a=5 also works for R # Python list ⇔ R vector: a = [1,3,5,7] ⇔ a <- c(1,3,5,7) a = [i for i in range(3,9)] ⇔ a <- c(3:9) # Python ‘for loop’: for val in [1,3,5]:
print(val)
# R ‘for loop’: for (val in c(1,3,5)){
print(val)
} # Python function: def new_function(a, b=5):
return a+b
# R function: new_function <- function(a, b=5) {
return (a+b)
} Inspecting dataframe # Python ⇔ R df.head() ⇔ head(df) df.head(3) ⇔ head(df,3) df.tail(3) ⇔ tail(df,3) df.shape[0] ⇔ nrow(df) df.shape[1] ⇔ ncol(df) df.shape ⇔ dim(df) df.info() ⇔ NO EQUIVALENT df.describe() ⇔ summary(df) # similar, not exactly the same NO EQUIVALENT ⇔ str(df) File I/O # Python import pandas as pd df = pd.read_csv(“input.csv”,
sep = “,”, header = 0)
df.to_csv(“output.csv”, index = False) # R df <- read.csv(“input.csv”,
header = TRUE, na.strings=c(“”,”NA”), sep = “,”)
write.csv(df, “output.csv”, row.names = FALSE) # na.strings: make sure NAs are not read as empty strings Create a new dataframe # Python import pandas as pd df = pd.DataFrame(dict(col_a=[‘a’,’b’,’c’], col_b=[1,2,3])) # R col_a <- c(‘a’,’b’,’c’) col_b <- c(1,2,3) df <- data.frame(col_a, col_b) Column / row filtering # Python: row filtering df[(df[‘column_1’] > 3) &
(df[‘column_2’].isnull())]
# R: row filtering df[(df$column_1 > 3) &
(is.na(df$column_2)), ]
OR library(dplyr) df %>% filter((column_1 > 3) & (is.na(column_2))) # Python ⇔ R: column filtering (keep columns) df[[‘c1’, ‘c2’]] ⇔ df[c(‘c1’, ‘c2’)] # OR: df[,c(‘c1’, ‘c2’)] # Python ⇔ R(with dplyr): column filtering (drop columns) df.drop([‘c1’, ‘c2’], axis=1) ⇔ df %>% select(-c(‘c1’, ‘c2’)) # Python ⇔ R: select columns by position df.iloc[:,2:5] ⇔ df[c(3:5)] # Note the indexing # Python: check if a column contains specific values df[df[‘c1’].isin([‘a’,’b’])] OR df.query(‘c1 in (“a”, “b”)’) # R: check if a column contains specific values df[df$c1 %in% c(‘a’, ‘b’), ] OR library(dplyr) df %>% filter(c1 %in% c(‘a’, ‘b’)) Missing value handling / count # Python: missing value imputation df[‘c1’] = df[‘c1’].fillna(0) OR df.fillna(value={‘c1’: 0}) # R: missing value imputation df$c1[is.na(df$c1)] <- 0 OR df$c1 = ifelse(is.na(df$c1) == TRUE, 0, df$c1) OR library(dplyr) library(tidyr) df %>% mutate(c1 = replace_na(c1, 0)) # Python ⇔ R: number of missing values in a column df[‘c1’].isnull().sum() ⇔ sum(is.na(df$c1)) Statistics for a single column # Python ⇔ R: count value frequency (Similar) df[‘c1’].value_counts() ⇔ table(df$c1) df[‘c1’].value_counts(dropna=False) ⇔ table(df$c1, useNA=’always’) df[‘c1’].value_counts(ascending=False) ⇔ sort(table(df$c1), decreasing = TRUE) # Python ⇔ R: unique columns (including missing values) df[‘c1’].unique() ⇔ unique(df$c1) len(df[‘c1’].unique()) ⇔ length(unique(df$c1)) # Python ⇔ R: column max / min / mean df[‘c1’].max() ⇔ max(df$c1, na.rm = TRUE) df[‘c1’].min() ⇔ min(df$c1, na.rm = TRUE) df[‘c1’].mean() ⇔ mean(df$c1, na.rm = TRUE) grouping and aggregations # Python: max / min / sum / mean / count tbl = df.groupby(‘c1’).agg({‘c2’:[‘max’, ‘min’, ‘sum’],
‘c3’:[‘mean’], ‘c1’:[‘count’]}).reset_index()
- tbl.columns = [‘c1’, ‘c2_max’, ‘c2_min’, ‘c2_sum’,
‘c3_mean’, ‘count’]
OR (for chained operations) tbl = df.groupby(‘c1’).agg(c2_max= (‘c2’, max),
c2_min= (‘c2’, min), c2_sum= (‘c2’, sum), c3_mean= (‘c2’, ‘mean’), count= (‘c1’, ‘count’)).reset_index()
# R: max / min / sum / mean / count library(dplyr) df %>% group_by(c1) %>%
- summarise(c2_max = max(c2, na.rm = T),
c2_min = min(c2, na.rm = T), c2_sum = sum(c2, na.rm = T), c3_mean = mean(c3, na.rm = T), count = n())
# Python: count distinct df.groupby(‘c1’)[‘c2’].nunique()
.reset_index().rename(columns={‘c2’:’c2_cnt_distinct’})
# R: count distinct library(dplyr) tbl <- df %>% group_by(c1)
%>% summarise(c2_cnt_distinct = n_distinct(c2))
creating new columns / altering existing columns # Python: rename columns df.rename(columns={‘old_col’: ‘new_col’}) # R: rename columns library(dplyr) df %>% rename(new_col = old_col) # Python: value mapping df[‘Sex’] = df[‘Sex’].map({‘male’:0, ‘female’:1}) # R: value mapping library(dplyr) df$Sex <- mapvalues(df$Sex,
from=c(‘male’, ‘female’), to=c(0,1))
# Python ⇔ R: change data type df[‘c1’] = df[‘c1’].astype(str) ⇔ df$c1 <- as.character(df$c1) df[‘c1’] = df[‘c1’].astype(int) ⇔ df$c1 <- as.integer(df$c1) df[‘c1’] = df[‘c1’].astype(float) ⇔ df$c1 <- as.numeric(df$c1) Updating column values by row filters # Python ⇔ R: df.loc[df[‘c1’]==’A’, ‘c2’] = 99 ⇔ df[df$c1==’A’, ‘c2’] <- 99 Joining / sorting # Python: inner join / left join import pandas as pd merged_df1 = pd.merge(df1, df2, on=’c1’, how=’inner’) merged_df2 = pd.merge(df1, df2, on=’c1’, how=’left’) OR (for chained operations) merged_df1 = df1.merge(df2, on=’c1’, how=’inner’) merged_df2 = df1.merge(df2, on=’c1’, how=’left’) # R: inner join / left join merged_df1 <- merge(x=df1,y=df2,by=’c1’) merged_df2 <- merge(x=df1,y=df2,by=’c1’,all.x=TRUE) OR library(dplyr) merged_df1 <- inner_join(x=df1,y=df2,by=’c1’) merged_df2 <- left_join(x=df1,y=df2,by=’c1’) # Python: sorting df.sort_values(by=[‘c1’,’c2’], ascending = [True, False]) # R: sorting library(dplyr) df %>% arrange(c1, desc(c2)) Concatenation / sampling # Python (import pandas as pd) ⇔ R: concatenation pd.concat([df1, df2, df3]) ⇔ rbind(df1, df2, df3) pd.concat([df1, df2], axis=1) ⇔ cbind(df1, df2) # Python random sample df.sample(n=3, random_state=42) # R random sample set.seed(42) sample_n(df, 3) An example of chained operations # Python: chained operations with ‘.’ df.drop(‘c1’, axis=1)
.sort_values(by=’c2’, ascending=False).assign(c3 = lambda x: x[‘c1’]*3 + 2).fillna(value={‘c2’: 0, ‘c4’:-99}).rename(columns={‘total’: ‘TOT’}).query(‘c3 > 10’)
# R: chained operations with ‘%>%’ library(dplyr) library(tidyr) df %>% select(-c(‘c1’)) %>%
arrange(desc(c2)) %>% mutate(c3 = c1*3 + 2) %>% mutate(c2 = replace_na(c2, 0),
c4 = replace_na(c4, -99)) %>%
rename(TOT = total) %>% filter(c3 > 10)