Splitting around an outer character in Perl. Minimal assumptions within group

Question

I am having a hard time adapting the answer in this thread to the following problem:

I would like to split the following string:

my $string = "foo{age}, bar{height}. something_else. baz{weight,so='yes',brothers=john.smith}.test{some}"

around the outer dots. The result should be an array holding

("foo{age}, bar{height}", 
 "foo{weight,parents='yes',brothers=john.smith}", 
 "test{some}")

I would like to avoid making assumptions about what's inside the groups inside {}.

How can I do this in Perl?

I tried adapting the following:

print join(",",split(/,\s*(?=\w+{[a-z,]+})/g, $string));

by replacing what's inside the character class [] without success.

Update:

The only characters not allowed within a {} group are { or }

The [solution](http://stackoverflow.com/a/17936618/1521179) for your other question still works here, you just have to change the separator: `/\G ($regex) [.] \s*/xg`. — amon, Aug 01 '13 at 01:08
@amon. GOT IT. Sorry I didn't quite understand the last part of your answer. I think I do now. Sorry for the potential duplicate. I will make that clear in the OP. — Amelio Vazquez-Reina, Aug 01 '13 at 01:14
Are your outside dots always exactly next to a closing `}` symbol? This would make it very easy, but I am assuming that it is just a coincidence of your example to look like that. — Joseph Myers, Aug 01 '13 at 01:19
@amon, actually just changing `[,;]` to `[.]` does not seem to do the trick. — Amelio Vazquez-Reina, Aug 01 '13 at 01:21
Thanks @JosephMyers. I clarified that. The dots don't need to be next to a closing `}`. — Amelio Vazquez-Reina, Aug 01 '13 at 01:22
Is it possible for you to have an empty field, i.e., two consecutive dots with nothing between them? This also affects how simple the solution can be. — Joseph Myers, Aug 01 '13 at 01:29
Never mind, I see that you have a working solution already. I had one running in a perl test file based on the same idea as my sentence suggested and in m.buettner's solution, but I wanted to know whether I could optimize it. — Joseph Myers, Aug 01 '13 at 01:33
@user815423426 Ah yes, I'm sorry. I wrote an answer that extends my previous approach to this problem, and I find it rather readable. However, m.buettner’s solution is quite nice. — amon, Aug 01 '13 at 01:38

score 5 · Accepted Answer · edited May 23 '17 at 12:05

5

Since you are not dealing with nested braces, the periods you want are those which are not "immediately" followed by a closing }. Where "immediately" means, without an opening { in between:

split(/[.]\s*(?![^{]*[}])/g, $string)

Alternatively, to match the parts you're interested in:

(?:[^.{}]|[{][^{}]*[}])+

Which can be "unrolled" to:

[^.{}]*(?:[{][^{}]*[}][^.{}]*)*

edited May 23 '17 at 12:05

Community

1
1

answered Aug 01 '13 at 01:09

Martin Ender

43,427
11
90
130

+1. Thanks. This is a good solution, but I updated the OP, since the dots dot not necessarily have to be next to a closing `}` – Amelio Vazquez-Reina Aug 01 '13 at 01:23
@user815423426 my solution does not assume that (in neither of the two approaches) – Martin Ender Aug 01 '13 at 01:24
Got it. I guess that you are right, since it's a weak assumption. I can probably work with this assumption, although the spirit of the post is to make *minimal assumptions* (effectly no assumptions) about what's inside `{}` – Amelio Vazquez-Reina Aug 01 '13 at 01:31
@m.buettner: forget my precedent comment, because I saw the pattern like this: `[^\s.{](?:[^{.]*{[^}]*})+` – Casimir et Hippolyte Aug 01 '13 at 01:37
1

Nice solution! Your split is great, although your other two are incorrect. E.g., "foo{age}, bar{height}. baz{weight,so='yes',brothers=john.smith}.test{some}" -> "foo", "", "age", "", ", bar", "", "height", "", "", " baz", "", "weight,so='yes',brothers=john", "", "smith", "", "", "test", "", "some", "", "" – Joseph Myers Aug 01 '13 at 01:37
1

@JosephMyers oh right, they were both missing a quantifier. fixed. – Martin Ender Aug 01 '13 at 10:17

score 1 · Answer 2 · answered Aug 01 '13 at 01:36

Here is how I would have solved the problem:

We define an item:
```
my $item = qr/ \w+ (?: [{] [^{}]* [}] )? /x;
```
That is, some word characters and optionally a section inside braces.
We define item groups, separated by comma:
```
my $item_group = qr/$item \s* (?: , \s* $item \s* )*/x;
```
That is, an $item followed by zero or more comma-item sequences.
We extract the results by matching for an item group that is followed by a period or the end of string:
```
my @result = $string =~ /\G ($item_group) \s* (?: [.] \s* | \z)/xg;
```

Output:

(
  "foo{age}, bar{height}",
  "something_else",
  "baz{weight,so='yes',brothers=john.smith}",
  "test{some}",
)

score 0 · Answer 3 · answered Aug 01 '13 at 01:09

0

You can do a match instead of a split, matching the alternatives of braces with anything between them (except braces) or non-braces with anything except for a dot.

answered Aug 01 '13 at 01:09

Joseph Myers

6,434
27
36

Splitting around an outer character in Perl. Minimal assumptions within group

Update:

3 Answers3