Re: [RFC/PATCH 1/9] ref-filter: add %(refname:lalignX) option
- From
Christian Couder <christian.couder@gmail.com>
- Date
- Jun 28, 2015, 04:53 UTC
- Message-ID
- <CAP8UFD2nnJYybMVPeQ5QMvsHmcqJf3tmAuSpfrCdqTaucKUuSw@mail.gmail.com>
- In-Reply-To
- <CAP8UFD1gT_2RKYc1vvBO9goGwPzFxhu9uMan5+cLO7eWxwwB_w@mail.gmail.com>
On Sat, Jun 27, 2015 at 10:02 PM, Christian Couder <christian.couder@gmail.com> wrote:
Show 25 quoted lines
> On Thu, Jun 25, 2015 at 1:43 PM, Karthik Nayak <karthik.188@gmail.com> wrote:
>
>> + if (starts_with(formatp, "lalign")) {
>> + const char *valp;
>> + int val;
>> +
>> + skip_prefix(formatp, "lalign", &valp);
>> + val = atoi(valp);
>
> After thinking about such code, I wonder if it would be better to
> support %(refname:lalign=X) instead of %(refname:lalignX).
>
> The reason why it might be interesting to require an = sign between
> "align" and the number X is that if we later want to introduce another
> option with a name that starts with "lalign", for example
> %(refname:lalignall=X) that would truncate the refname if it is bigger
> than X), we might be more backward compatible with old git versions
> that implement %(refname:lalign=X) but not %(refname:lalignall=X).
>
> We will be more backward compatible because the above call to
> starts_with() would probably be something like:
>
> if (starts_with(formatp, "lalign=")) {
>
> which means that old git versions would ignore something like "lalignall=X".Another reason is that it would be simpler if we ever want to have arbitrary string parameters, like %(refname:substitute=%/%\%).