diazrenata / psinet-template-docs Goto Github PK

Downloading "User-facing" as an excel doc preserves validation but breaks the Date columns, because Excel doesn't like the YYYYMMDD date format and displays it as a bunch of hashtags #########.

In the event we have folks who want to use Excel, we could a) try and fix this by using a different date format b) create our own Excel template doc and send it to them POST A LINK TO IT!

I've realized that it may become unwieldly to use Google Sheets for distributing the spreadsheet and collecting answers, because we'd have to create a new Google Sheet for everyone (undermining the nice collaboration/shared link features).

So I suggest keeping the validation logic simple, and continuing our development in Google Sheets. Before we distribute it, download it, convert it to Excel, and fix any broken logic.

• using 2 packages to connect to excel. One preserves the column names as they are written, which I was initially using to write back properly. But you don't need to write the column names at all, and so it may be simpler to use just one.
• late in the script I figured out to use bind_rows with 0 row table to fill in missing/not applicable columns with NAs when you don't know which columns those will be. may want to propagate that on up.
• I think it would be good/cover all bases to create toy data (or a modification of the provided data) where the pl_codes do link up, because - if I understand Rafa correctly - that is expected to be the norm, and that differs in a couple of substantive ways from the example I have.
• if the pl_codes do match, there will then be 2 sources of data on things like species, treatment. Do we hard-code choosing one over the other, or do we compare and if they are different force a user to make a decision?
• there's also an edge(?) case where some, but not all, of the pl_codes match up.
• What, exactly, is the difference between pl_name and pl_code? I think these scripts should always use pl_code because those are assigned by SAPFLUXNET....right?
• turn into a series of functions? I don't like hundreds-of-line scripts, and it would be good to take out the hardcoding paths.
• does this eventually see life as an interactive script c.f. the portal rodent cleaning script?
• partway through the script I converted from creating new dataframes with the values hard-coded-transferred over from sfn_* to the template-styled data frame; then I figured out how to do it using mutate and renames. Revise the earlier code to take that approach? (For some of the earlier tables - ones with just 1 row - that may make no sense, but I want to check if it does for others).
• more generally it is pretty fiddly (and feels like playing with fate) to have such determinedly non-machine-readable column names in the PSInet template. Perhaps there is a way to make this more robust, e.g. by having a table linking the human and machine friendly tables, and then swapping them right away?