RegularParser = Uncaught Error: Maximum function nesting level of '256' reached, aborting!

[rhukster]

We're using thunderer/shortcode in our Grav CMS shortcode-core plugin: https://github.com/getgrav/grav-plugin-shortcode-core

And it was reported to me that while using also using the Gantry5 Plugin while Shortcode Core was active, resulted in an infinite loop bug. I dug into this a bit and it appears that some HTML code output by the plugin causes this PHP infinite loop bug when using the RegularParser. It works fine with the RegexParser or WordpressParser BTW.

I have created a little test project where I'm using JUST the thunderer/shortcode library and the sample content in question, and it's definitely repeatable even out of the context of Grav and the Shortcode Core plugin.

shortcode-infinite-loop.zip

In this file, you can see the HTML that is causing the issue in question. I've tried 'cleaning' the HTML, but that does not change the results.

My hunch is that it's the HTML data attributes that are throwing the parser for a loop:

...
<li>
    <a
        data-g5-ajaxify
        data-g5-ajaxify-target="[data-g5-content]"
        href="/testing/grav-g5/admin/gantry/positions?nonce=85998f180935f47b4b01ad293ac1325c"
        ><i class="fa fa-fw fa-object-group" aria-hidden="true"></i>
        Positions</a
    >
</li>
<li>
    <a
        data-g5-ajaxify
        data-g5-ajaxify-target="[data-g5-content]"
        href="/testing/grav-g5/admin/gantry/menu?nonce=85998f180935f47b4b01ad293ac1325c"
        ><i class="fa fa-fw fa-bars" aria-hidden="true"></i>
        <span>Menu</span></a
    >
</li>
...

I'm inclined to disable the Grav Shotcode Core while Gantry5 plugin is processing, as it's really not being used here, it's just that Grav is processing ALL page content for potential shortcodes.

[thunderer]

Hi @rhukster, thanks for your report, please look at #72. Can you verify my results using dev-parser-improvements branch?

Your test data runs in ~20ms and uses around 10MiBs of memory in my local environment, Blackfire reports 80ms and ~9MiBs, baseline master ran in 12.5s and used 514MiBs of memory. All tests pass so I'm pretty sure it works correctly, but let me know how does it look on your side.

If everything is fine we have over 100x performance gain on time and memory and RegularParser can be used in all situations without any penalty.

[rhukster]

I can confirm that I've tested the parser-improvements branch and the test case I added as part of #72 now works, an also the actual issue as reported to in getgrav/grav-plugin-shortcode-core#53 and gantry/gantry5#2426 have also been resolved!

I notice now that when there are no actual shortcodes, the regular parser is indeed very fast to process the content, in fact, the fastest of all the 3 parsers.

I've not done any comprehensive performance testing on the parsers for actual shortcode work, but it definitely is faster, and I can find no visible issues with my custom shortcodes page I use for testing.

Nice work! 👍

[thunderer]

@rhukster During the verification phase, I managed to reproduce #70 and added the fix to #72. Could you please verify the results again? I promise I'm not doing anything else with the PR, so it will be merged once I get your approval.

Unfortunately, the fix reduces the performance gains I mentioned above, but we're still well above 10x improvement in both time and memory (850ms and 47MiB on Blackfire, 660ms and 46MiB on local environment, for your test data vs the same baseline). Details are described in the PR if you're interested.

[rhukster]

Please see my comments on the PR itself. Alas, the issue has returned with the latest commit, but does seem to be related to xdebug being on. (was this way with stable release too, I just didn't realize it before).

[rhukster]

Update... latest commits in #72 branch again fixes the issue for me.

[thunderer]

If this PR fixed the issue, may I ask you to close it? I will tag the next version soon, I just need to review the commits since the last version and prepare some release notes.

[rhukster]

Ya all good for me...

[rhukster]

Any word on timing for a new release?

[thunderer]

@rhukster I'm very happy to announce that v0.7.0 was just tagged, you can find the release notes here. While I was writing them I found a lot of small things which could be improved, so #73 will probably become v0.7.1 in the near future.

Thanks a lot for using Shortcode!