70 Investigate full coverage for tcplDefine. Updated tcplQuery/tcplDefine for example data

[cthunes]

Removed the check for driver type within tcplDefine and moved it to tcplQuery. tcplQuery accepts a SQL select statement from tcplDefine and can translate it for use with example .rda data using sqldf.

tcplDefine tests are passing and now provide 100% coverage of the function. Closes #70.

[cthunes]

Github actions check failing one test (passing locally) from test-tcplDefine, which is unrelated to the function properly functioning, and a bug with expect_snapshot. This inconsistency will be resolved in a different PR

[cthunes]

See PR #75 for resolution to failing test

[brown-jason]

Once we merge the other branches that fix the checks we can rerun the github actions and confirm we are seeing 100% test coverage here.

[cthunes]

Checks are passing with 100% coverage. They are also all passing locally. However, the warning that was appearing here in the Github check as a fail before is now appearing here as WARN result. Looks like this has nothing to do with the snapshot testing anymore, but the warning is probably something we still want to address? Tracing back to sqldf. Perhaps it's 'stringsAsFactors = F' part? Maybe that should be set to true? What do you think @brown-jason?

══ Warnings ════════════════════════════════════════════════════════════════════
── Warning ('test-tcplDefine.R:3:3'): empty returns full dictionary ────────────
no DISPLAY variable so Tk is not available
Backtrace:
     ▆
  1. ├─testthat::expect_snapshot_value(tcplDefine(), style = "serialize") at test-tcplDefine.R:3:2
  2. │ ├─testthat:::with_is_snapshotting(force(x))
  3. │ └─base::force(x)
  4. └─tcpl::tcplDefine()
  5.   └─tcpl::tcplQuery(query = qstring, db = getOption("TCPL_DB"), tbl = tbl)
  6.     ├─data.table::as.data.table(sqldf(query, stringsAsFactors = F))
  7.     └─sqldf::sqldf(query, stringsAsFactors = F)
  8.       └─base::requireNamespace("tcltk", quietly = TRUE)
  9.         ├─base::tryCatch(loadNamespace(package, ...), error = function(e) e)
 10.         │ └─base (local) tryCatchList(expr, classes, parentenv, handlers)
 11.         │   └─base (local) tryCatchOne(expr, names, parentenv, handlers[[1L]])
 12.         │     └─base (local) doTryCatch(return(expr), name, parentenv, handler)
 13.         └─base::loadNamespace(package, ...)
 14.           └─base (local) runHook(".onLoad", env, package.lib, package)
 15.             ├─base::tryCatch(fun(libname, pkgname), error = identity)
 16.             │ └─base (local) tryCatchList(expr, classes, parentenv, handlers)
 17.             │   └─base (local) tryCatchOne(expr, names, parentenv, handlers[[1L]])
 18.             │     └─base (local) doTryCatch(return(expr), name, parentenv, handler)
 19.             └─tcltk (local) fun(libname, pkgname)

[brown-jason]

Would something like this fix the issue?
forestgeo/learn#186

[cthunes]

Would something like this fix the issue? forestgeo/learn#186

Investigation has been mostly unsuccessful. This link is for users of travis CI, rather than github actions. The closest I could get to travis'

before_install:
- export DISPLAY=:99.0

is github actions':

env:
  DISPLAY: 99.0

No longer getting the same warning that there is no display variable, but instead getting a warning that it "couldn't connect to display "99". I tried 0.0 as well with the same result. Any other ideas?