You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@trafficserver.apache.org by James Peach <jp...@apache.org> on 2018/02/01 16:24:54 UTC
Re: Proposed new ts api calls
> On Jan 30, 2018, at 9:15 AM, Walt Karas <wk...@oath.com.INVALID> wrote:
>
> Revised:
>
> const char *TSRemapFromUrlStringGet(const TSHttpTxn txnp, int &length);
>
> const char *TSRemapToUrlStringGet(const TSHttpTxn txnp, int &length);
>
> These would return, from the transaction state object,
> url_map.getFromURL() and url_map.getToURL() respectively, in strong
> format. A null pointer would be returned if the values were empty.
> The length of the string is returned in 'length'. (Nul-termination
> should not be expected/presumed.) . The plugin must release the memory
> for the string by calling TSFree().
This is C, so no reference parameters. Following the existing TSAPI conventions, I'd recommend
TSReturnCode TSRemapFromUrlGet(TSHttpTxn txnp, TSMBuffer bufp, TSMLoc obj);
char * TSRemapFromUrlStringGet(TSHttpTxn txnp, int *length);
TSReturnCode TSRemapToUrlGet(TSHttpTxn txnp, TSMBuffer bufp, TSMLoc obj);
char * TSRemapToUrlStringGet(TSHttpTxn txnp, int *length);
In the documentation, please be sure to specify which hooks these APIs may be called from. Consider writing an example plugin that can be consumed by tests and included in the documentation.
I'm a not that fazed by the StringGet() APIs proposed here, but we should consider whether we want to continue this API pattern, or defer the stringification to TSUrlStringGet().
> On Mon, Jan 29, 2018 at 4:55 PM, Walt Karas <wk...@oath.com> wrote:
>> The plugin would own them and would have to call TSFree() when done
>> with them. They would be nul terminated.
>>
>> On Mon, Jan 29, 2018 at 4:52 PM, Alan Carroll
>> <so...@oath.com.invalid> wrote:
>>> What are the ownership properties of the memory indicated by the returned
>>> pointers? Are these strings guaranteed to be null terminated or is a length
>>> needed?
>>>
>>> On Mon, Jan 29, 2018 at 3:53 PM, Shu Kit Chan <ch...@gmail.com> wrote:
>>>
>>>> What does the "I" stand for in "TSIRemapFromUrlStringGet" ?
>>>>
>>>> On Mon, Jan 29, 2018 at 10:02 AM, Walt Karas <wk...@oath.com.invalid>
>>>> wrote:
>>>>> const char *TSIRemapFromUrlStringGet(const TSHttpTxn txnp);
>>>>>
>>>>> const char *TSIRemapToUrlStringGet(const TSHttpTxn txnp);
>>>>>
>>>>> These would return, from the transaction state object,
>>>>> url_map.getFromURL() and url_map.getToURL() respectively, in strong
>>>>> format. A null pointer would be returned if the values were empty.
>>>>
Re: Proposed new ts api calls
Posted by Alan Carroll <so...@oath.com.INVALID>.
Looks mostly good to me, except a minor coding issue. The API documentation
is on that PR as well, anyone who is concerned should look at it. I think
it avoids the issues that have been brought up on this exchange.
On Tue, Feb 20, 2018 at 5:52 PM, Walt Karas <wk...@oath.com.invalid> wrote:
> I've issued a pull request for the implementation of this:
>
> https://github.com/apache/trafficserver/pull/3145
>
> On Thu, Feb 1, 2018 at 12:49 PM, Leif Hedstrom <zw...@apache.org> wrote:
> >
> >
> >> On Feb 1, 2018, at 9:35 AM, Alan Carroll <so...@oath.com.INVALID>
> wrote:
> >>
> >> That's a good suggestion, although shouldn't it be
> >>
> >> TSReturnCode TSRemapFromUrlGet(TSHttpTxn txnp, TSMBuffer bufp, TSMLoc*
> obj);
> >>
> >> That is the TSMLoc should be an out parameter, not an in. In that case
> I'd
> >> drop the ...StringGet() and use the generic TSUrlStringGet() if a
> string is
> >> needed.
> >
> >
> > I’m definitely not convinced we need those additional StringGet() APIs,
> when the existing one would definitely work. As for the unused marshaling
> buffer, we can look into that for 8.0.0, and decide if it’s never needed,
> and clean that up and fix the APIs accordingly. I’d much rather prefer
> that, that introduce new, duplicated functionality APIs.
> >
> > — Leif
> >
>
Re: Proposed new ts api calls
Posted by Walt Karas <wk...@oath.com.INVALID>.
I've issued a pull request for the implementation of this:
https://github.com/apache/trafficserver/pull/3145
On Thu, Feb 1, 2018 at 12:49 PM, Leif Hedstrom <zw...@apache.org> wrote:
>
>
>> On Feb 1, 2018, at 9:35 AM, Alan Carroll <so...@oath.com.INVALID> wrote:
>>
>> That's a good suggestion, although shouldn't it be
>>
>> TSReturnCode TSRemapFromUrlGet(TSHttpTxn txnp, TSMBuffer bufp, TSMLoc* obj);
>>
>> That is the TSMLoc should be an out parameter, not an in. In that case I'd
>> drop the ...StringGet() and use the generic TSUrlStringGet() if a string is
>> needed.
>
>
> I’m definitely not convinced we need those additional StringGet() APIs, when the existing one would definitely work. As for the unused marshaling buffer, we can look into that for 8.0.0, and decide if it’s never needed, and clean that up and fix the APIs accordingly. I’d much rather prefer that, that introduce new, duplicated functionality APIs.
>
> — Leif
>
Re: Proposed new ts api calls
Posted by Leif Hedstrom <zw...@apache.org>.
> On Feb 1, 2018, at 9:35 AM, Alan Carroll <so...@oath.com.INVALID> wrote:
>
> That's a good suggestion, although shouldn't it be
>
> TSReturnCode TSRemapFromUrlGet(TSHttpTxn txnp, TSMBuffer bufp, TSMLoc* obj);
>
> That is the TSMLoc should be an out parameter, not an in. In that case I'd
> drop the ...StringGet() and use the generic TSUrlStringGet() if a string is
> needed.
I’m definitely not convinced we need those additional StringGet() APIs, when the existing one would definitely work. As for the unused marshaling buffer, we can look into that for 8.0.0, and decide if it’s never needed, and clean that up and fix the APIs accordingly. I’d much rather prefer that, that introduce new, duplicated functionality APIs.
— Leif
Re: Proposed new ts api calls
Posted by Walt Karas <wk...@oath.com.INVALID>.
When calling this, where would the value for the bufp parameter come
from? How would it be used in the implementation?
On Thu, Feb 1, 2018 at 10:35 AM, Alan Carroll
<so...@oath.com.invalid> wrote:
> That's a good suggestion, although shouldn't it be
>
> TSReturnCode TSRemapFromUrlGet(TSHttpTxn txnp, TSMBuffer bufp, TSMLoc* obj);
>
> That is the TSMLoc should be an out parameter, not an in. In that case I'd
> drop the ...StringGet() and use the generic TSUrlStringGet() if a string is
> needed.
Re: Proposed new ts api calls
Posted by Alan Carroll <so...@oath.com.INVALID>.
That's a good suggestion, although shouldn't it be
TSReturnCode TSRemapFromUrlGet(TSHttpTxn txnp, TSMBuffer bufp, TSMLoc* obj);
That is the TSMLoc should be an out parameter, not an in. In that case I'd
drop the ...StringGet() and use the generic TSUrlStringGet() if a string is
needed.