Combine timetype wiki twitter

[nmdefries]

pvt_twitter and pub_wiki can take dates in either day or week format. These are provided as separate, mutually exclusive (will fail) params dates and epiweeks. This format disagrees with the interface for pub_covidcast, which also supports days and weeks. Change the twitter and wiki interfaces to match, with new params time_type and time_values.

This also makes supporting the date range default to wildcard "*" (all dates) easier -- if both dates and epiweeks params were set to "*" by default, the user would have to specify which date type they wanted to use anyway, a la time_type, since we wouldn't be able to automatically determine precedence.

[pullflow-com]

nmdefries (GitHub user) ▸ 🔘 Marked ready for review

[pullflow-com]

nmdefries (GitHub user) ▸ 🔍 Review requested from dshemetov, dsweber2 and brookslogan

[pullflow-com]

nmdefries (GitHub user) ▸ 🔍 Review requested from dshemetov, dsweber2 and brookslogan

[pullflow-com]

nmdefries (GitHub user) ▸ 🔍 Review requested from dshemetov, dsweber2 and brookslogan

[dsweber2]

The tests cover what I would think of. Only 1 minor note.

I'm a bit surprised that these weren't following the convention of the other endpoints. Looking for other exceptions, I didn't find any, thankfully. Good catch!

• CDC matches the api in using epiweeks
• pub_covid_hosp_state_timeseries uses dates, matching the API
• pub_delphi does actually use epiweeks only
• same for pub_dengue_nowcast, pvt_dengue_sensors, pub_ecdc_ili, pub_flusurv, pub_fluview_clinical, pub_fluview, pub_gft, pub_kcdc_ili, pub_nidss_dengue, and pub_nidss_flu
• I'm not actually sure about pvt_norostat. The docs have ellipses covering the date.
• pub_nowcast, as epiweeks matches their api docs
• pub_paho_dengue has ellipses covering the date
• pvt_quidel, pvt_sensors, use epiweeks in both

[dsweber2]

I'm not entirely sure this is necessary, or else pub_covidcast is missing something important, since it doesn't do this. I would maybe make a function for this bit of logic if it is necessary, since it (should?) be included 3 times.

(I mean this and if statements below)

[capnrefsmmat]

Here it's necessary because the default value is c("day", "week"), but you can only have one time_type. match.arg() takes the first if the default is not overridden. It's a common convention in R that if the argument can be one of several string values, you make the default list all of them, and match.arg() takes care of picking the first (if none is provided) or ensuring the provided value is valid otherwise.

pub_covidcast() doesn't set a default for this argument, so it's not strictly necessary. However, pub_covidcast() could do match.arg(time_type, c("day", "week")) to enforce that the type must be one or the other but not both.

[pullflow-com]

dsweber2 ▸ ⎘ R/endpoints.R: I'm not entirely sure this is necessary, or else pub_covidcast is missing something important, since it doesn't do this. I would maybe make a function for this bit of logic if it is necessary, since it (should?) be included 3 times.

(I mean this and if statements below)