Keep the epi_df class when head and tail are used

[rachlobay]

As per issue 96 head.epi_df and tail.epi_df to keep the epi_df class when they are used. Note the tests for this in test-methods-epi_df.R

[brookslogan]

The comments above are very minor. This PR could be merged to address #96 without any changes.

[dajmcdon]

@rachlobay I spent a bit of time looking into this, and I think we're doing this the wrong way (my fault). head() actually calls utils:::head.matrix() eventually, which uses [.tbl_df. So rather than implementing head.epi_df() we should actually be implementing [.epi_df. Those changes will then propagate to a call to head() or tail() as well as keeping the class if someone does some_epi_df[1:5, ].

I think we want it to drop the class and the metadata if they do column selection: some_epi_df[ ,1:2] should just go to a tbl_df.

Good examples are in tibble or probably even better tsibble. I suspect that we essentially need most of the code in the tSibble link.

[rachlobay]

Ok. I have removed head.epi_df and tail.epi_df and instead implemented [.epi_df based on the tsibble [.tbl_ts. The tests are passing for head() and tail(). @dajmcdon please have a look at the [.epi_df when you can and let me know if you would like me to make any changes.

[dajmcdon]

I think there are a few additional tests that we need here, otherwise, the rest is linting.

[dajmcdon]

I'm not positive, but I suspect that this doesn't actually work. This is just giving you the names without the values.

I think you can safely do additional_metadata = att_x$additional_metadata but you should probably add a test to check that it works.

[brookslogan]

Some notes:

• Above test failures are related to the additional_metadata; at least some [might be] due to a missing[/extra] $additional_metadata in the tests.
• The additional_metadata can be set directly in as_epi_df; see pkgload::dev_help("as_epi_df"). [EDIT: or not... I just tried this out and the additional_metadata are not forwarded; list() is passed to new_epi_df instead. Maybe a function signature copy-paste bug.]
• Either additional_key_name and "additional_key_type" should be changed to other_keys and "<some column that acts as part of the key in addition to geo and time values>" or turned into something without "key" in the name. E.g., source_url -> "https://example.com/data.csv" or something like that.

[rachlobay]

@brookslogan Thanks for the comments! To let you know, I ended up fixing the bug that you spotted in as_epi_df() and used that to add the other key (as in %>% as_epi_df(additional_metadata = c(other_keys = "indicator_var"))).