[libvirt] [PATCH 07/11] vsh: Introduce complete command

Michal Privoznik mprivozn at redhat.com
Wed Nov 8 15:00:35 UTC 2017


On 11/08/2017 03:23 PM, Martin Kletzander wrote:
> On Tue, Nov 07, 2017 at 01:22:55PM +0100, Michal Privoznik wrote:
>> This command is going to be called from bash completion script in
>> the following form:
>>
>>  virsh complete "start --domain"
>>
>> Its only purpose is to return list of possible strings for
>> completion. Note that this is a 'hidden', unlisted command and
>> therefore there's no documentation to it.
>>
>> Signed-off-by: Michal Privoznik <mprivozn at redhat.com>
>> ---
>> tools/virsh.c      |  1 +
>> tools/virt-admin.c |  1 +
>> tools/vsh.c        | 68
>> ++++++++++++++++++++++++++++++++++++++++++++++++++++++
>> tools/vsh.h        | 14 +++++++++++
>> 4 files changed, 84 insertions(+)
>>
>> diff --git a/tools/virsh.c b/tools/virsh.c
>> index 7d6dc2620..f830331f6 100644
>> --- a/tools/virsh.c
>> +++ b/tools/virsh.c
>> @@ -846,6 +846,7 @@ static const vshCmdDef virshCmds[] = {
>>     VSH_CMD_PWD,
>>     VSH_CMD_QUIT,
>>     VSH_CMD_SELF_TEST,
>> +    VSH_CMD_COMPLETE,
>>     {.name = "connect",
>>      .handler = cmdConnect,
>>      .opts = opts_connect,
>> diff --git a/tools/virt-admin.c b/tools/virt-admin.c
>> index 5d7ef7988..c24ed95c0 100644
>> --- a/tools/virt-admin.c
>> +++ b/tools/virt-admin.c
>> @@ -1356,6 +1356,7 @@ static const vshCmdDef vshAdmCmds[] = {
>>     VSH_CMD_PWD,
>>     VSH_CMD_QUIT,
>>     VSH_CMD_SELF_TEST,
>> +    VSH_CMD_COMPLETE,
>>     {.name = "uri",
>>      .handler = cmdURI,
>>      .opts = NULL,
>> diff --git a/tools/vsh.c b/tools/vsh.c
>> index 121669574..136acb0ab 100644
>> --- a/tools/vsh.c
>> +++ b/tools/vsh.c
>> @@ -3419,3 +3419,71 @@ cmdSelfTest(vshControl *ctl ATTRIBUTE_UNUSED,
>>
>>     return true;
>> }
>> +
>> +/* ----------------------
>> + * Autocompletion command
>> + * ---------------------- */
>> +
>> +const vshCmdOptDef opts_complete[] = {
>> +    {.name = "string",
>> +     .type = VSH_OT_ARGV,
>> +     .flags = VSH_OFLAG_EMPTY_OK,
>> +     .help = N_("partial string to autocomplete")
>> +    },
>> +    {.name = NULL}
>> +};
>> +
>> +const vshCmdInfo info_complete[] = {
>> +    {.name = "help",
>> +     .data = N_("internal command for autocompletion")
>> +    },
>> +    {.name = "desc",
>> +     .data = N_("internal use only")
>> +    },
>> +    {.name = NULL}
>> +};
>> +
>> +bool
>> +cmdComplete(vshControl *ctl, const vshCmd *cmd)
>> +{
>> +    bool ret = false;
>> +#ifdef WITH_READLINE
>> +    const vshClientHooks *hooks = ctl->hooks;
>> +    int stdin_fileno = STDIN_FILENO;
>> +    const char *arg = NULL;
>> +    char **matches = NULL, *tmp = NULL, **iter;
>> +
>> +    if (vshCommandOptStringQuiet(ctl, cmd, "string", &arg) <= 0)
>> +        goto cleanup;
>> +
>> +    /* This command is flagged VSH_CMD_FLAG_NOCONNECT because we
>> +     * need to prevent auth hooks reading any input. Therefore we
>> +     * have to close stdin and then connect ourselves. */
>> +    VIR_FORCE_CLOSE(stdin_fileno);
>> +
>> +    if (!(hooks && hooks->connHandler && hooks->connHandler(ctl, true)))
>> +        goto cleanup;
>> +
>> +    autoCompleteOpaque = ctl;
>> +
>> +    rl_basic_word_break_characters = " \t\n\\`@$><=;|&{(";
>> +    if (VIR_STRDUP(rl_line_buffer, arg) < 0)
>> +        goto cleanup;
>> +
>> +    while ((tmp = strpbrk(arg, rl_basic_word_break_characters)))
>> +        arg = tmp + 1;
>> +
>> +    if (!(matches = vshReadlineCompletion(arg, 0, 0)))
>> +        goto cleanup;
>> +
>> +    for (iter = matches; *iter; iter++)
>> +        printf("%s\n", *iter);
>> +
>> +    ret = true;
>> + cleanup:
>> +    for (iter = matches; iter && *iter; iter++)
>> +        VIR_FREE(*iter);
>> +    VIR_FREE(matches);
>> +#endif /* WITH_READLINE */
> 
> Do we really need readline for it?  Did I miss something or are we
> really just using it here just so we don't have to split the
> vshReadlineParse() in some way?  Don't get me wrong, I'm OK with that,
> the split would be too complicated and who doesn't use readline, right?
> It just feels weird.

Yes we do need readline unfortunately. vshReadlineCompletion() calls
rl_completion_matches() which filters strings returned by
vshReadlineParse() so that list of strings presented to user corresponds
to their input. We could reimplement the rl_* function ourselves if we
really want to be readline independent. On the other hand - readline is
everywhere, so why should we bother?

Michal




More information about the libvir-list mailing list