Zero-length error when trying to Lag xts time series?

Question

I have basic xts object generated using the following.

library(quantmod)
temp1 <- getSymbols("GOOG",src = 'yahoo',from=Sys.Date()-50,to = Sys.Date(),auto.assign=FALSE)
temp2 <- temp1$GOOG.Close
head(temp2)
           GOOG.Close
2020-05-20    1406.72
2020-05-21    1402.80
2020-05-22    1410.42
2020-05-26    1417.02
2020-05-27    1417.84
2020-05-28    1416.73
lag(temp2,1)
Error in c.xts(NA_real_, c(1406.719971, 1402.800049, 1410.420044, 1417.02002,  : 
  zero-length vectors with non-zero-length index are not allowed

As far as I can tell temp2 is not zero length so I don't understand the error.

Similarly the diff(temp2) function does work as expected.

This seems to have started when I recently updated to R 4.0.1

I cannot find any explanation for this error considering I do have a non-zero vector.

Answer 1

Do you have dplyr loaded, by chance? That's the only way I can replicate your error.

dplyr masks the stats::lag() generic function that ships with base R. That breaks method dispatch for all S3 classes that define a lag() method. So that means lag() may or may not work for those classes.

R> head(lag(temp2, 1))
           GOOG.Close
2020-05-26         NA
2020-05-27    1417.02
2020-05-28    1417.84
2020-05-29    1416.73
2020-06-01    1428.92
2020-06-02    1431.82
R> suppressPackageStartupMessages(library(dplyr))
R> head(lag(temp2, 1))
Error in c.xts(NA_real_, c(1417.02002, 1417.839966, 1416.72998, 1428.920044,  : 
  zero-length vectors with non-zero-length index are not allowed

So you either need to: 1) not load dplyr, or 2) explicitly call stats::lag() everywhere you don't want to use dplyr::lag().