Correct the Spread command in xbrl_get_data( )

[Karnner2]

Hi! There may have been a change to the SEC reporting of the XBRL. I spent some time correcting the code so it works.

In the XBRL_get_data() (called by XBRL_get_statements function) you have a set of contextId and elementId filters. Brilliant work in getting this to adapt the correct filters from the underlying taxonomy -- this is truly a work of art in how you managed it but! Spread is prone to error because it fails if there are duplicate elements with different row ids (which there are if there is a new row in the financial statements). This is a catastrophic error -- I say catastrophic because the program virtually fails at random over various years.

Please remove the dyplr::spread__(elementId, fact ) command and replace it with the following:

reshape2::dcast(formula = contextId + 
                  startDate + endDate + decimals ~ elementId, value.var = "fact", 
                fun.aggregate = mean)

While I have chosen fun.aggregate = mean, this does not actually affect anything and I have validated that it does the correct pull. Sum also seems to work -- but I figured if there was some sort of aberrant issue it would be better to average than sum.

The reshape2 will l do a specific long-to-wide conversion where Spread fails. We can see in various R templates that reshape2 is the obsidian blade to Spread's steel surgeon's knife with respect to directness. I've tested it and it seems to be a general correction to the problems of the package. You can simply paste it where the spread_ command is and it should work fine.

I am also forking this project because I am getting tired of doing trace('xbrl_get_data', edit=TRUE) every time I load, R. Hope you don't mind! But I hope you will consider the adjustment so I can download directly from your repository in the future!

All the best, JTFJ