onefile_to_egor fragile to column ordering

[mbojan]

onefile_to_egor seems to be very fragile to the column order in the input data frame. Consider the following input data frame which is two rows from GSS 2004:

dd <- read.csv(textConnection(  '"id","numgiven","sex","sex1","sex2","sex3","sex4","sex5","close12","close13","close14","close15","close23","close24","close25","close34","close35","close45"
10,6,1,2,2,1,2,2,3,1,2,2,2,2,2,1,1,1
36,6,2,2,2,2,1,1,2,2,2,1,2,2,2,1,2,2'
), as.is=TRUE)

I can create egor object from that with

e <- with(
  dd,
  onefile_to_egor(
    egos = dd,
    pmin(numgiven,5), 
    ID.vars=list(ego="id"), 
    attr.start.col="sex1",
    attr.end.col="sex5",
    max.alters=5,
    aa.first.var="close12", 
    aa.regex="^(?<attr>[[:alpha:]]+)(?<src>[[:digit:]])(?<tgt>[[:digit:]])$"
  )
)

but it will fail if I add a variable at the end:

# add a variable at the end
dd$foobar <- 1

e <- with(
  dd,
  onefile_to_egor(
    egos = dd,
    pmin(numgiven,5), 
    ID.vars=list(ego="id"), 
    attr.start.col="sex1",
    attr.end.col="sex5",
    max.alters=5,
    aa.first.var="close12", 
    aa.regex="^(?<attr>[[:alpha:]]+)(?<src>[[:digit:]])(?<tgt>[[:digit:]])$"
  )
)

which gives

Error in `[.data.frame`(e.wide, -nm) : object 'nm' not found

It seems that the functions is interpreting the new foobar column as giving information on alter-alter ties.

I'm not yet sure what's the best way to address it, probably one of:

• Write in the documentation that the input data frame has to have variables and blocks of variables in specific order (e.g. alter-alter variables last)
• Modify the function so that the user will need to specify variable names for all alter and alter-alter variable block explicitly. E.g. something like f( aattrs=list(sex = "sex[0-9]"), aaties = list(close = "close[0-9]{2}") ) ....
• Give up on trying to support every possible variation in which ppl may store egocentric data and provide constructor functions only for, say, (1) long format and (2) separate dataframes...

[krivit]

Maybe just have aa.start.col and aa.end.col arguments, with aa.end.col defaulting to the last column in the data frame?