duckplyr 0.4.0

Features

• Use built-in rfuns extension to implement equality and inequality operators, improve translation for as.integer(), NA and %in% (#83, #154, #148, #155, #159, #160).
• Reexport non-deprecated dplyr functions (#144, #163).
• library(duckplyr) calls methods_overwrite() (#164).
• Only allow constant patterns in grepl().
• Explicitly reject calls with named arguments for now.
• Reduce default memory limit to 1 GB.

Bug fixes

• Stricter type checks in the set operations intersect(), setdiff(), symdiff(), union(), and union_all() (#169).
• Distinguish between constant NA and those used in an expression (#157).
• head(-1) forwards to the default implementation (#131, #156).
• Fix cli syntax for internal error message (#151).
• More careful detection of row names in data frame.
• Always check roundtrip for timestamp columns.
• left_join() and other join functions call auto_copy().
• Only reset expression depth if it has been set before.
• Require fallback if the result contains duplicate column names when ignoring case.
• row_number() returns integer.
• is.na(NaN) is TRUE.
• summarise(count = n(), count = n()) creates only one column named count.
• Correct wording in instructions for enabling fallback logging (@TimTaylor, #141).

Chore

• Remove styler dependency (#137, #138).
• Avoid error from stats collection.

Documentation

• Mention wildcards to read multiple files in ?df_from_file (@andreranza, #133, #134).

Testing

• Reenable tests that now run successfully (#166).
• Synchronize tests (#153).
• Test that vec_ptype() does not materialize (#149).
• Improve telemetry tests.
• Promote equality checks to expect_identical() to capture differences between doubles and integers.