Skip to content

RecordSeparators

Status: Experimental

present and registered, but lightly documented and not yet covered by dedicated tests.

Description

RecordSeparators

is an option for Read and ReadList giving the list of strings that separate records. The default is {"\r\n", "\n", "\r"}.

Examples (2)

Every input below was run against the current Mathilda build and its output recorded.

Applications (2)

Split records on commas

In[1]:= s = OpenWrite["/tmp/mathilda_rec.txt"]; WriteString[s, "a,b,,c"]; Close[s]; ReadList["/tmp/mathilda_rec.txt", Record, RecordSeparators -> {","}]
Out[1]= {"a", "b", "c"}

Keep the empty field

In[2]:= s = OpenWrite["/tmp/mathilda_rec.txt"]; WriteString[s, "a,b,,c"]; Close[s]; ReadList["/tmp/mathilda_rec.txt", Record, RecordSeparators -> {","}, NullRecords -> True]
Out[2]= {"a", "b", "", "c"}

Implementation notes

Definition. RecordSeparators is an option for Read and ReadList giving the list of strings that delimit records (for the Record read type). It has no builtin and no value of its own ā€” it is an inert, Protected keyword carried in the reader's option list, with default {"\r\n", "\n", "\r"}.

Representation. A bare EXPR_SYMBOL (interned SYM_RecordSeparators). Both read.c and readlist.c seed their default option set with rule_of(SYM_RecordSeparators, {"\r\n", "\n", "\r"}); a user RecordSeparators -> {...} rule overrides it. The reader's next_record scanner splits on any of these strings. The companion NullRecords option decides whether an empty field between two adjacent separators is kept.

Usage & limits. Protected; meaningful only as a Read/ReadList option. Because it has a registered default, SetOptions[Read, RecordSeparators -> ...] works. Related tokenisation options are WordSeparators, TokenWords, NullRecords and NullWords.

Attributes: none registered.

References

Notes & additional examples

Notes

RecordSeparators is an option for Read and ReadList giving the strings that delimit records (used with the Record read type). The default is {"\r\n", "\n", "\r"} ā€” one record per line.

It pairs with NullRecords: by default an empty field between two adjacent separators is dropped, so "a,b,,c" split on "," yields {"a", "b", "c"}; with NullRecords -> True the empty field is kept, giving {"a", "b", "", "c"}. RecordSeparators is an inert, Protected keyword with a registered default, so SetOptions[Read, RecordSeparators -> ...] works. The word-level analogues are WordSeparators and TokenWords.