g.part5.definedays warning: "number of items to replace is not a multiple of replacement length"

[l-k-]

(I'm not very familiar with part 5 code, so I was hoping someone might know right away the correct way to fix this. If not, I can give it a shot next week.)

I ran GGIR master branch on the UK biobank sample cwa file, found here: biobank.ctsu.ox.ac.uk/ukb/ukb/examples/accsamp.zip (This is a file use by profileGGIR)

When part 5 is executed, I get the following warning message:

In breaks_i[bi] = fullQqq[grep(breaks[bi], ts$time[fullQqq])] :
  number of items to replace is not a multiple of replacement length

The exact command I used:

GGIR(datadir="/Users/user/cwa", outputdir="/Users/user/res", overwrite=TRUE, do.parallel = FALSE, do.cal=FALSE, printsummary=TRUE, verbose = TRUE, windowsizes = c(10,660,1800))

The warning is triggered in g.part5.definedays, here, for the very first time block.

This very first time block spans the time from 2013-07-13 10:11:00 to 2013-07-20 09:55:50, so its first midnight is 2013-07-14, but this midnight is not recorded in nightsi. So the very first fullQqq window contains the time span from 2013-07-13 10:11:00 all the way to 2013-07-14 23:59:50, which means that it contains not one but two "23:59:50" timestamps.

So fullQqq[grep(breaks[2], ts$time[fullQqq])] matches not one but two timestamps. And then breaks_i[2] = fullQqq[grep(breaks[2], ts$time[fullQqq])] results in the warning "number of items to replace is not a multiple of replacement length"

You can see the details in the screenshot below:

There are two ways to solve this --

either

1. make sure that the first midnight 2013-07-14 is recorded in nightsi,
   or
2. modify the assignment breaks_i[bi] = fullQqq[grep(breaks[bi], ts$time[fullQqq])] to take the last of the fullQqq[grep(breaks[bi], ts$time[fullQqq])] elements.

I'm not sure which one of these solutions is preferable, since I don't know a lot about part 5.

Additional info

1. Sensor brand: Axivity
2. Data format: cwa
3. Are you using a sleep diary to guide the sleep detection: NO
4. Copy of R command used: GGIR(datadir="/Users/user/cwa", outputdir="/Users/user/res", overwrite=TRUE, do.parallel = FALSE, do.cal=FALSE, printsummary=TRUE, verbose = TRUE, windowsizes = c(10,660,1800))
5. Have you tried processing your data based on GGIR's default argument values? Does the issue you report still appear? YES, and I got the same warning message, but I didn't re-do my debugging with default values.

Desktop (please complete the following information):

• OS: Mac OS
• GGIR Version: 3.0-5, latest master

Before submitting

• Have you tried the steps to reproduce? Do they include all relevant data and configuration? Does the issue you report still appear there?
• Have you tried this on the latest master branch from GitHub?

[vincentvanhees]

Thanks for creating this issue. I see the problem is that the ts$time[fullQqq] spans more than 24 hours, because qqq defaults to going from 1 to the first midnight, while the first 2 midnights are not present. I am now preparing a PR to catch this.