qualiaa / column-grep Goto Github PK

User should be able to negate a match. Ideally, this should work and feel consistent for any index, range or name.

Probably the nicest way to do this is following awk, and allowing a leading ! before a particular match value to negate it. For example, /^type-/,!3:5 will match any column starting with "type-" except columns 3-5.

There is an open question whether these should override each other in order of appearance; for now if a match is negated then it should override any other match.

Currently, column-grep assumes that the input is a CSV. However, following many other tools, most notably awk, it should allow the user to describe both input and output separators for records and fields.

Long names should be descriptive:

• -f --input-field-separator
• -F --output-field-separator
• -r --input-record-separator
• -R --output-record-separator

If only the input version is provided for a particular option, the output should default to the provided value.

Options to do input and output from files rather than standard streams.

• -i<file> --input <file>: read input from file, disable stdin
• -o<file> --output <file>: output to file, disable stdout (allow multiple times)

If <file> is - then it in interpreted as the appropriate standard stream.

Currently the output is rebuilt from fields and records using ByteString.intercalate - the documentation highlights that this is slow and that ByteString.Builder is 20% faster.

All regex are currently case-sensitive. PCRE has an compilation flag for case-insensitivity, we should expose this as a flag; following grep:

• -i, --ignore-case for case-insensitive matching
• --no-ignore-case for case-sensitive (default).

Could also follow GNU sed with flags on the end of regexes: /re/i for case-insensitive, /re/I for case-sensitive.

The following compilation options from pcre_compile are available through Text.Regex.PCRE.CompOpts and should be toggle-able:

PCRE_UTF8 and PCRE_UNGREEDY should be on by default; PCRE_CASELESS pertains to #2 .

PCRE_NEWLINE_* and PCRE_MULTILINE pertain to #5, as certain values of field and record separators modify their interpretation.

Currently, the order in which output-specs appear is ignored, and duplicate columns are ignored. This is a sensible default, but it would be nice to support intentional column duplication and output-ordering.

For example, the following would produce column 1 multiple times:

column-grep --duplicates 1 1 1

This would produce columns 1, then 3, then 2:

column-grep --preserve-order 1 3 2

This would produce 1, then 3, then 2, then 3 again:

column-grep --preserve-order 1 3 2 3

If --no-headers is given, do not treat the first line specially. Produce an error for any attempts to match by name.

Shorthand could be -n?