spectro_analysis Error: "integer overflow"

[CarolineTroy]

Hi, I'm trying to run the spectro_analysis() function on Mac for a data table of about 5000 selections made by auto_detec() from bird audio files.

It is getting about halfway through before giving this error: "Error in m[(fl[1]:fl[2]) + 1, ] : subscript out of bounds"
When I look at the warnings, there are fifty "In (0:(n - 1)) * f : NAs produced by integer overflow" errors.

Any advice on how to resolve this would be much appreciated!

[maRce10]

can you please share the code? You can also try check_sels() on the selection table to try to find problematic selections.

[CarolineTroy]

Hi, thank you for your reply.

Below is the code for the detection process and the analysis:
XCWarbler1_detec = auto_detec(bp = c(.5, 7), threshold = 5, mindur = 0.4, maxdur = 1.3, ssmooth = 800)
XCWarbler1_detec2 = na.omit(XCWarbler1_detec)
XCW1_Analysis = spectro_analysis(XCWarbler1_detec2, bp = c(.5, 8))

When I try checksels() on the selection table after removing NA's, I get "all selections are OK but not all sound files have the same sampling rate (potentially problematic, particularly for cross_correlation())"

[maRce10]

Sounds like a bug. Can you please share the data in extended selection format? You will have to convert the data frame into that format, then save it as an RDS file and send it to me (upload it to this link). Like this:

# convert into extended selection table
XCWarbler1_detec2_est <- selection_table(XCWarbler1_detec2, extended = TRUE, confirm.extended = FALSE)


# save RDS
saveRDS(XCWarbler1_detec2_est, "XCWarbler1_detec2_est.RDS")