Closes #3664 -- diligently avoid partial argument matching

[MichaelChirico]

#3664

Confirmed we're doing OK for attr with:

grep -r "\battr[(]" R | grep -v "exact=TRUE"

There's one call to attr that's not using exact = TRUE inside of .shallow, since it's actually being called as attr<- (didn't look so at first glance but it threw "too many arguments" error when I first tried).

A bunch of partial matches were being used in test.data.table(). With the commit here, test.data.table() runs fine using the options suggested by @vlulla.

[jangorecki]

generally looks good. Although we are losing tests for partial matching-ness. I seen quite commonly use of id instead of id.vars or fun instead of fun.aggregate. If in future we will add new argument that would break this partial matching, then tests would detect such "breaking change".

[MichaelChirico]

i thought of that and convinced myself we didn't want such tests but maybe you're right. If so, though, we should create tests explicitly for that purpose. (and skip them if the warnPartialMatch options are set, I guess)

[codecov]

Codecov Report

Merging #3676 into master will not change coverage.
The diff coverage is 100%.

@@           Coverage Diff           @@
##           master    #3676   +/-   ##
=======================================
  Coverage   98.25%   98.25%           
=======================================
  Files          69       69           
  Lines       13084    13084           
=======================================
  Hits        12856    12856           
  Misses        228      228

Impacted Files	Coverage Δ
R/test.data.table.R	100% <ø> (ø)	⬆️
R/print.data.table.R	95.37% <100%> (ø)	⬆️
R/fread.R	100% <100%> (ø)	⬆️
R/IDateTime.R	98.75% <100%> (ø)	⬆️
R/data.table.R	97.87% <100%> (ø)	⬆️
R/duplicated.R	100% <100%> (ø)	⬆️
R/setkey.R	98.7% <100%> (ø)	⬆️
R/foverlaps.R	98.1% <100%> (ø)	⬆️
R/fcast.R	99.4% <100%> (ø)	⬆️
R/frank.R	100% <100%> (ø)	⬆️
... and 3 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c3b8b2f...ca9e4d2. Read the comment docs.

[vlulla]

@MichaelChirico you did a thorough job of fixing lots of files! Great work! And, thank you!

[mattdowle]

Very nice PR - thanks @vlulla and @MichaelChirico! I just did a minor tweak to the news item and turned these options on so the tests always run with them on, to avoid any future partial argument matching creeping back in. One reveal on nanotime: eddelbuettel/nanotime#49
Tweeted here: https://twitter.com/MattDowle/status/1149868351662903296

[MichaelChirico]

great! I meant to suggest the same for turning on the option. onwards and upwards!