Set.Seed with R 2.15.2

set.seed with R 2.15.2

set.seed() reinitializes the random number generator.

set.seed(12345)
rnorm(5)
[1]  0.5855288  0.7094660 -0.1093033 -0.4534972  0.6058875

set.seed(12345)
rnorm(5)
[1]  0.5855288  0.7094660 -0.1093033 -0.4534972  0.6058875

set.seed(12345)
rnorm(5)
[1]  0.5855288  0.7094660 -0.1093033 -0.4534972  0.6058875

R: Strange behavior while saving list() with save() from function output

look at

> attr(ok[[1]]$terms,".Environment")
<environment: 0x9bcf3f8>
> attr(ok2[[1]]$terms,".Environment")
<environment: R_GlobalEnv>

also

> ls(envir = attr(ok[[1]]$terms,".Environment"))
[1] "i"  "k"  "tt"

so ok is dragging around the environment of the function with it.

Also read ?object.size

 The calculation is of the size of the object, and excludes the
 space needed to store its name in the symbol table.

 Associated space (e.g. the environment of a function and what the
 pointer in a ‘EXTPTRSXP’ points to) is not included in the
 calculation.

For example define a test2 and an ok3

test2 = function(k){
    tt = vector('list',k)
    for(i in 1:k) tt[[i]] = lm(a0~b1+b2+b3,data = data)
    rr = tt
    tt
}

ok3 <- test2(2)
save(ok3, 'ok3.RdData')

> file.info('ok3.RData')$size
[1] 5043933
> file.info('ok.RData')$size
[1] 3366005
> file.info('ok2.RData')$size
[1] 1678851

> ls(envir = attr(ok3[[1]]$terms,".Environment"))
[1] "i"  "k"  "rr" "tt"

so ok is roughly twice as big as ok2 because it has the extra tt and ok3 is three times as big as it has tt and rr

> c(object.size(ok),object.size(ok2),object.size(ok3))
[1] 4019336 4019336 4019336

There is related discussion here

table() generating NAs when there are no NAs in the underlying data

After installing the data.table package and doing some preliminaries...

require(data.table)
n0<- 1e5
n <- 1e6
DT <- data.table(A1 = sample(1:n0, n, replace = TRUE),B1 = sample(1:n0, n, replace = TRUE))

this does the trick.

setkey(DT,A1)
DT[
    DT[,.N,by=A1],
    countC:=N
]

When you access a data.table with DT[i,j], you can select rows with i and do something else with j, just like in data.frames.

DT[,.N,by=A1] selects all rows (since i is blank) and counts rows for each "A1" using the special variable .N.

After setting column "A1" as key for DT, we can pass a data.table -- in this case DT[,.N,by=A1] -- in i to merge back the information in the latter data.table. In j, we create a new column in DT using countC:=N. The three vignettes on data.table's CRAN page are a good place to start learning more about how this works.

The question at hand. Oh, I think I see what the original problem was. Suppose unique(x)=c(1,2,4). If you try table(x)[x], you will be trying to access table(x)[1], table(x)[2] and table(x)[4]. The last one is undefined since the length of the table is only 3. R always returns NA when we access indices greater than the length of a vector. For example, look at (1:3)[4].

In your case, if you are missing any unique values in 1:n0 that are not at the very top, you will see NAs.

Assignment by reference with sum() in data.table() yields incorrect result

This is not a problem with data.table, but rather, human error ;)

To replicate, here is some sample data. I've included some NA values to see the results of the sum function with and without the argument to remove NAs, which is na.rm, not na.remove:

set.seed(1)
test <- data.table(Year = rep("Y1", 15),
                   ID = c(rep(210, 9), rep(3197, 6)),
                   Count = sample(c(0, 1, NA), 15, 
                                  prob=c(.2, .65, .15), 
                                  replace=TRUE),
                   key = "Year,ID")
test
#     Year   ID Count
#  1:   Y1  210     1
#  2:   Y1  210     1
#  3:   Y1  210     1
#  4:   Y1  210    NA
#  5:   Y1  210     1
#  6:   Y1  210    NA
#  7:   Y1  210    NA
#  8:   Y1  210     0
#  9:   Y1  210     1
# 10:   Y1 3197     1
# 11:   Y1 3197     1
# 12:   Y1 3197     1
# 13:   Y1 3197     0
# 14:   Y1 3197     1
# 15:   Y1 3197     0

Before we create our new column, let's just do some aggregation to see what happens with the different options for sum.

test[, list(annualCount = sum(Count)), by = key(test)]
#    Year   ID annualCount
# 1:   Y1  210          NA
# 2:   Y1 3197           4
test[, list(annualCount = sum(Count, na.rm = TRUE)), by = key(test)]
#    Year   ID annualCount
# 1:   Y1  210           5
# 2:   Y1 3197           4

Now, create your new column, with the results you expected.

test[, annualCount := sum(Count, na.rm = TRUE), by = key(test)][]
#     Year   ID Count annualCount
#  1:   Y1  210     1           5
#  2:   Y1  210     1           5
#  3:   Y1  210     1           5
#  4:   Y1  210    NA           5
#  5:   Y1  210     1           5
#  6:   Y1  210    NA           5
#  7:   Y1  210    NA           5
#  8:   Y1  210     0           5
#  9:   Y1  210     1           5
# 10:   Y1 3197     1           4
# 11:   Y1 3197     1           4
# 12:   Y1 3197     1           4
# 13:   Y1 3197     0           4
# 14:   Y1 3197     1           4
# 15:   Y1 3197     0           4

Quickly reading very large tables as dataframes

An update, several years later

This answer is old, and R has moved on. Tweaking read.table to run a bit faster has precious little benefit. Your options are:

1. Using vroom from the tidyverse package vroom for importing data from csv/tab-delimited files directly into an R tibble. See Hector's answer.

2. Using fread in data.table for importing data from csv/tab-delimited files directly into R. See mnel's answer.

3. Using read_table in readr (on CRAN from April 2015). This works much like fread above. The readme in the link explains the difference between the two functions (readr currently claims to be "1.5-2x slower" than data.table::fread).

4. read.csv.raw from iotools provides a third option for quickly reading CSV files.

5. Trying to store as much data as you can in databases rather than flat files. (As well as being a better permanent storage medium, data is passed to and from R in a binary format, which is faster.) read.csv.sql in the sqldf package, as described in JD Long's answer, imports data into a temporary SQLite database and then reads it into R. See also: the RODBC package, and the reverse depends section of the DBI package page. MonetDB.R gives you a data type that pretends to be a data frame but is really a MonetDB underneath, increasing performance. Import data with its monetdb.read.csv function. dplyr allows you to work directly with data stored in several types of database.

6. Storing data in binary formats can also be useful for improving performance. Use saveRDS/readRDS (see below), the h5 or rhdf5 packages for HDF5 format, or write_fst/read_fst from the fst package.

---

The original answer

There are a couple of simple things to try, whether you use read.table or scan.

1. Set nrows=the number of records in your data (nmax in scan).

2. Make sure that comment.char="" to turn off interpretation of comments.

3. Explicitly define the classes of each column using colClasses in read.table.

4. Setting multi.line=FALSE may also improve performance in scan.

If none of these thing work, then use one of the profiling packages to determine which lines are slowing things down. Perhaps you can write a cut down version of read.table based on the results.

The other alternative is filtering your data before you read it into R.

Or, if the problem is that you have to read it in regularly, then use these methods to read the data in once, then save the data frame as a binary blob with save saveRDS, then next time you can retrieve it faster with load readRDS.

---

Related Topics