Johannes,

Thanks a lot, I modified the extension carefully according to all of your
notes. Everything has been covered by tests, here they are:
https://github.com/DmitryKoterov/dom_varimport/tree/master/tests

About your note "You should add the dom extension as dependency in the
module structure, this can improve error messages in some cases when using
shared libraries ...". I thought that THERE IS the dependency declared, see
https://github.com/DmitryKoterov/dom_varimport/blob/master/config.m4#L25 .
I also added DOM extension dependency directly to PHPT tests as well -
please see e.g.
https://github.com/DmitryKoterov/dom_varimport/blob/master/tests/002_formats.phpt#L6

Maybe you've meant some other dependencies? How to declare them? I googled,
but I failed.



On Mon, Apr 21, 2014 at 6:47 PM, Johannes Schlüter
<[email protected]>wrote:

> Hi,
>
> On Fri, 2014-04-18 at 17:11 +0400, Dmitry Koterov wrote:
> > Dom_varimport is a simple PHP extension to convert nested arrays into
> > DOMDocument with minimum CPU and time consumption.
>
> > Samples:
> > https://github.com/DmitryKoterov/dom_varimport
>
> From a quick review this seems to be fine, a few minor comments:
>
>       * You should remove the [RM][INIT|SHUTDOWN] functions, especially
>         the ones with R lead to unneeded calls and pointer dereferencing
>         etc. on every request
>       * You should add the dom extension as dependency in the module
>         structure, this can improve error messages in some cases when
>         using shared libraries ...
>       * The declaration of "static int le_dom_varimport;" makes me
>         suspicious - when compiling with proper error messages enabled
>         this should trigger a warning. Maybe there are other warnings
>         from the compiler (CFLAGS=-Wall etc.)?
>       * In general we prefer php_error_docref over php_error. This could
>         (depending on the user's configuration) link directly to the
>         docs from the error message :-)
>       * In php_dom_varimport() for the default case I suggest throwing
>         an error ("Unknown type") to be on the safe side.
>       * From a design I don't think the fallback to "item" for "invalid"
>         tag names is a good choice - I think an error might be more
>         appropriate. Ideally the "key" attribute name could also be
>         configured, especially when thinking about XML namespaces
>
> All minor things and the last thing is more a matter of taste :)
>
> johannes
>
>
>

Reply via email to