Add write_dwc() function

[peterdesmet]

Implements #256

@jreubens @jonasmortelmansvliz @damianooldoni @jdpye @sarahcd here's a first version of the write_dwc() function that transforms acoustic telemetry data to Darwin Core that can be harvested by OBIS and GBIF. It follows the recommendations we discussed, which are based on conventions introduced for Movebank data in the movepub write_dwc() function. An eml.xml file is not created.

You can test it by installing:

devtools::install_github("inbo/etn#257")

@jdpye @sarahcd this will work on LifeWatch RStudio Server only for now, but you can see the SQL file in this PR to see the transformation.

Then:

library(etn)
con <- connect_to_etn()
write_dwc(
  animal_project_code = "2014_demer" # or other project
  # optional parameters are: directory, rights_holder, license

If all goes well, a dwc_occurrence.csv will be written to disk.

[peterdesmet]

Note: dataGeneralizations currently has the fixed value subsampled by hour. The goal is to make this subsampled by hour: first of 3 record(s) cf. Movebank.

Calculating the number of records per group is however much slower than the current approach. Do you consider it worth to provide this information at the cost of performance? Update: currently blocked by #259.

Using the function on project 2013_albertkanaal which has 6 million records currently takes 39 secs.

[peterdesmet]

For those who cannot run the function, here's the DwC file dwc_occurrence.csv for the following query:

write_dwc(animal_project_code = "2014_demer", rights_holder = "INBO")

[peterdesmet]

Regarding #257 (comment):

Count per group is now included as subsampled by hour: first of 6 record(s) (cf. Movebank). Duplicates are excluded. The function takes the same amount of time as previously (35 seconds for a big project like 2013_albertkanaal).

[peterdesmet]

Only a couple of pending questions to finalize this: #256 (comment)

[peterdesmet]

All Darwin Core mapping issues are resolved (#256).

@jreubens @jonasmortelmansvliz @jdpye note that the current implementation only exports an Occurrence core (cf. Movebank data), not an Extended measurement or fact extension. The latter could be added in the future, but it should then be discussed what to express in there (scope potentially huge).

This PR is pending

• Technical review by @PietrH
• Minor version bump
• Rebuilding of website

[PietrH]

Before merging I suggest the following changes:

• replace expect_equal with expect_identical where appropriate
• Allow write_dwc() to return objects rather than files

The latter similar to inbo/camtraptor#181

[peterdesmet]

@PietrH all comments addressed.

[PietrH]

Looks good to me.

[sarahcd]

Super belatedly, I checked out the sql and example output dwc_occurrence (without access to the LifeWatch server), it all looks sensible to me. Happy to coordinate with ETN at any point on aligning future changes with Movebank. Thank you for all the work @peterdesmet !