Re: [hackers] [PATCH 1/3][sbase] sort: Fix handling of -k n,n case

From: Richard Ipsum <richardipsum_AT_vx21.xyz>
Date: Wed, 1 Jan 2020 01:36:41 +0100

On Tue, Dec 31, 2019 at 02:35:57PM -0800, Michael Forney wrote:
> On 2019-12-31, Richard Ipsum <richardipsum_AT_vx21.xyz> wrote:
> > When field_start is equal to field_end use field N as the sort key.
>
> I think this is not just a problem with `-k n,n`, but any time
> field_end is specified. Only up to the beginning of the end field is
> compared (if there are more fields after it).

True

>
> I think we should keep the invariant that a->data + a->len is the end
> of the line, even though end.len isn't used after the call to
> skipcolumn(&end, 0).
>
> What do you think about the following diff instead?
>
> diff --git a/sort.c b/sort.c
> index dfc383f..a51997f 100644
> --- a/sort.c
> +++ b/sort.c
> _AT_@ -66,11 +66,10 @@ skipcolumn(struct line *a, int skip_to_next_col)
>
> if (fieldsep) {
> if ((s = memmem(a->data, a->len, fieldsep, fieldseplen))) {
> - if (skip_to_next_col) {
> + if (skip_to_next_col)
> s += fieldseplen;
> - a->len -= s - a->data;
> - a->data = s;
> - }
> + a->len -= s - a->data;
> + a->data = s;
> } else {
> a->data += a->len - 1;
> a->len = 1;
>

Looks good

Thanks,
Richard
Received on Wed Jan 01 2020 - 01:36:41 CET