Inside survey_stat_factor(), "reverse" conversion of peel variable back from numeric to character

[bschneidr]

This is one way to fix #78 and #74.

Inside survey_stat_factor(), after .svy$variables[[peel_var]] is converted to a character vector, the corresponding column in the output out[[peel_var]] is converted back to the original type. In conducting the conversion from numeric to character, the format() function is used to maintain the maximum possible precision when converting from numeric to character.

My only qualm with this approach is that it is possible that some desired level of numeric precision is lost when converting from numeric to character, and perhaps there is a similarly simple solution which would avoid this issue of precision.

[gergness]

seems reasonable, can you add tests though?

[bschneidr]

Good call.

It turns out something is going wrong when using double numeric grouping variables with many decimal places. The precision is preserved in the output, but sometimes the estimates and standard errors come back as NA. I'm still trying to figure out why this happens. Edit: Please disregard this- I had messed something up in my local environment when I wrote that.

[codecov-commenter]

Codecov Report

Merging #88 into master will increase coverage by 0.31%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master      #88      +/-   ##
==========================================
+ Coverage   79.94%   80.25%   +0.31%     
==========================================
  Files          20       20              
  Lines        1022     1028       +6     
==========================================
+ Hits          817      825       +8     
+ Misses        205      203       -2     

Impacted Files	Coverage Δ
R/survey_statistics.r	92.73% <100.00%> (+1.06%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6c379d9...8121cea. Read the comment docs.

[bschneidr]

I've added three tests which I think should cover everything:

1. survey_count()/survey_tally() work with an integer grouping variable
2. survey_count()/survey_tally() work with a double grouping variable
3. When a double grouping variable is used, the values of that variable in the output from survey_count()/survey_tally() are within .Machine$double.eps of the values in the input dataset.

With the addition of the third test, I think it should be ok to remove the warning message: warning("Coercing ", peel_name, " to character.", call. = FALSE):

srvyr/R/survey_statistics.r

Line 743 in a9c86e5

warning("Coercing ", peel_name, " to character.", call. = FALSE)

[gergness]

Thank you!