Allan Caffee <allan.caffee@xxxxxxxxx> writes:
> I assumed that the +1 in your example was a typo since AFAIKS ARRAY_SIZE
> should give us one past the last index.
You are correct.
> Also if git is to be expanded allow the use of non-ANSI color codes (or
> already does so) the strbuf_escape_sequence_length needs to be updated
> to accept the relevant escape codes.
Actually, I am starting to hate this.
Just step back a bit and imagine how you would do this, if you _were_
writing an application to do this kind of thing, generating output
directly to the terminal. You obviously would not seek back and count the
width of what you sent out. Instead,...?
That's right. You just keep a running total of how much you sent, iow,
what column you expect the current cursor should be. Can't we do the same
thing here?
> @@ -312,6 +351,33 @@ static struct commit_list *first_interesting_parent(struct git_graph *graph)
> return next_interesting_parent(graph, parents);
> }
>
> +unsigned short graph_get_current_column_color(const struct git_graph *graph)
static?
> +static void graph_increment_column_color(struct git_graph *graph)
> +{
> + graph->default_column_color = (graph->default_column_color + 1) %
> + ARRAY_SIZE(column_colors);
COLUMN_COLORS_MAX?
> +unsigned short graph_find_commit_color(const struct git_graph *graph,
> + const struct commit *commit)
> +{
static?
> @@ -714,10 +790,30 @@ static void graph_output_commit_char(struct git_graph *graph, struct strbuf *sb)
> strbuf_addch(sb, '*');
> }
>
> +void graph_draw_octopus_merge(const struct git_graph *graph,
> + struct strbuf *sb)
static?
> @@ -789,6 +880,17 @@ static void graph_output_commit_line(struct git_graph *graph, struct strbuf *sb)
> graph_update_state(graph, GRAPH_COLLAPSING);
> }
>
> +struct column *find_new_column_by_commit(struct git_graph *graph,
> + struct commit *commit)
> +{
static?
> diff --git a/strbuf.c b/strbuf.c
> index a884960..666460d 100644
> --- a/strbuf.c
> +++ b/strbuf.c
> @@ -1,3 +1,4 @@
> +#include <ctype.h>
> #include "cache.h"
> #include "refs.h"
BAD. Do not directly include system headers. If you need isgraph(),
please support it as part of sane_ctype. But if you count what you
emitted so far, you would not have to do this at all.
> +/*
> + * Return the length of the escape sequence in a string buffer
> + * starting at index i. If there is no escape sequence starting at
> + * return 0.
> + */
> +size_t strbuf_esc_sequence_length(const struct strbuf *sb, size_t i)
> +{
> + size_t start = i;
> + if (sb->buf[i] != '\033')
> + return 0;
> + ++i;
> +
> + if (i >= sb->len || sb->buf[i] != '[')
> + return 0;
> + ++i;
> + while (i < sb->len && isdigit(sb->buf[i]))
> + ++i;
> + if (i >= sb->len || sb->buf[i] != 'm')
> + return 0;
These preincrements are extremely unreadable at least for me.
if (sb->buf[i++] != '\033')
return 0;
if (sb->len <= i || sb->buf[i++] != '[')
return 0;
...
But again the point is hopefully moot.
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html