NEWS.md

In siftr: Fuzzily Search a Dataframe to Find Relevant Columns

Github-installed in-development version (> 1.1.0)

siftr 1.1.0

• FIXED

  • Removed useHash = TRUE from sample.int() inside should_approx() to match changes made in R 4.2.0. sample.int() now decided for itself whether to useHash or not. Closes https://github.com/DesiQuintans/siftr/issues/16.

• ADDED

  • save_dictionary() writes an empty exclude column to factor files.

• CHANGED

  • Changes to sift()'s announcement of a dictionary's contents. Closes https://github.com/DesiQuintans/siftr/issues/15:
    • Now says "{dataframe name} contains n columns..." instead of the generic "Dictionary contains n columns...".
    • Now shows head and tail of column names rather than just the head.
  • save_dictionary() writes only one cell of the ordered column, to conform to what real-world users would do.

siftr 1.0.1

• ADDED

  • User is now warned if .dist is used with an orderless search (because .dist is ignored in those cases).

• FIXED

  • Numeric vectors with NAs give a correct peek of their non-NA contents, rather than being reported as NA entirely.

• CHANGED

  • Suggestion to increase .dist in the case of no matches is now hidden for orderless searches.

siftr 1.0.0

• Renamed package to siftr to avoid name collision with existing sift package on CRAN that I somehow missed.

• Initial CRAN submission.

• ADDED

  • options_sift() gets a new option: sift_peeklength. This controls the approximate length of the rand_unique entries in the data dictionary, i.e. a list of unique values in each column. This "full peek" is used as part of the "haystack" that actually gets searched by sift(). It defaults to 3000 characters, but the final length increases when separators are added. Previously, a length limit of only 500 characters was hard-coded in. 3000 characters is about the length of a 1-page Word document at default settings.

• FIXED

  • Fixed a bug where columns that had multiple class() (e.g. "labelled" and "integer") would create a dictionary with two entries per dataframe column.
  • has_class() can deal with multi-classed variables now.
  • Row names are now discarded from generated dictionaries.

• CHANGED

  • some_uniques() has short-circuit routes for datatypes that don't need the full "random sampling to get a list of its unique values" treatment. So far this is: Factors, Logicals, and Numerics.
  • Changed "peek" separator to vertical bar | from comma , because some data may use commas within string values.
  • save_dictionary() generates factor files for each unique factor in the dataframe now, according to the tsv2label spec.
  • Sample data (mtcars_lab) now has a list column added.
  • should_approx() now uses sample.int() with the useHash argument, which performed better than sample().

sift 0.2.0 (2023-03-27)

• ADDED

  • save_dictionary() allows you to save a data dictionary in a form that my other package, tsv2label, will accept. Closes #11.
  • options_sift() prints the status of all options when invoked with no arguments. Closes #12.

• FIXED

  • sift() returns a dataframe of all results, not just the first n = sift_limit results.

sift 0.1.0 (2023-03-18)

• Initial commit.

siftr documentation built on July 26, 2023, 5:07 p.m.