Re: Avoid startup overhead by saving compiled large XSD for Xerces-J validation

--000000000000f1e44c05de5afe1c
Content-Type: text/plain; charset="UTF-8"

Created https://issues.apache.org/jira/browse/XERCESJ-1745

On Fri, May 6, 2022 at 1:53 AM Mukul Gandhi <[email protected]> wrote:

> Hi Mike,
>     Answering this question (my personal opinions), while 2.12.3
> release discussion is in progress in another thread :)
>
> On Thu, May 5, 2022 at 2:09 AM Mike Beckerle <[email protected]> wrote:
> > I imagine that to validate XML, Xerces does something akin to
> "compiling" the XSD into lower-level data structures for faster use when
> actually parsing (and validating) the incoming XML being parsed.
> >
> > Question 1: Is that true?
>
> Yes, Xerces compiles/translates XSD syntax to low-level data
> structures. I think that, Xerces compiling the XSD syntax into
> low-level data structures, is there to adhere to XSD defined component
> models of various XSD syntax elements, more than for any performance
> reasons.
>
> > Question 2: Is it possible to get this "compilation" of the large XSD
> schema done, and then serialize the resulting java object to a file, and
> reload this pre-compiled thing so as not to face this compiling overhead at
> startup time?
>
> I find this is a good idea.
>
> Although currently not implemented, within Xerces, I think we could
> have an api to serialize and deserialize org.apache.xerces.xs.XSModel,
> or even javax.xml.validation.Schema to a file (using java's
> java.io.Serializable mechanism). I think, currently
> org.apache.xerces.xs.XSModel (and many of its dependencies) etc does
> not implement the interface java.io.Serializable, which we need to do
> so, if we need to implement this feature.
>
> We also would need to make performance measurements, if in case, we'd
> implement XSD schema serialization feature.
>
> If possibly, you could create a jira issue (new feature) for this
> requirement, that can make this requirement easy to track.
>
>
> --
> Regards,
> Mukul Gandhi
>

--000000000000f1e44c05de5afe1c
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Created=C2=A0<a href=3D"https://issues.apache.org/jira/bro=
wse/XERCESJ-1745">https://issues.apache.org/jira/browse/XERCESJ-1745</a></d=
iv><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On =
Fri, May 6, 2022 at 1:53 AM Mukul Gandhi &lt;<a href=3D"mailto:mukulg@apach=
e.org">[email protected]</a>&gt; wrote:<br></div><blockquote class=3D"gmail=
_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204=
,204);padding-left:1ex">Hi Mike,<br>
=C2=A0 =C2=A0 Answering this question (my personal opinions), while 2.12.3<=
br>
release discussion is in progress in another thread :)<br>
<br>
On Thu, May 5, 2022 at 2:09 AM Mike Beckerle &lt;<a href=3D"mailto:mbeckerl=
[email protected]" target=3D"_blank">[email protected]</a>&gt; wrote:<br>
&gt; I imagine that to validate XML, Xerces does something akin to &quot;co=
mpiling&quot; the XSD into lower-level data structures for faster use when =
actually parsing (and validating) the incoming XML being parsed.<br>
&gt;<br>
&gt; Question 1: Is that true?<br>
<br>
Yes, Xerces compiles/translates XSD syntax to low-level data<br>
structures. I think that, Xerces compiling the XSD syntax into<br>
low-level data structures, is there to adhere to XSD defined component<br>
models of various XSD syntax elements, more than for any performance<br>
reasons.<br>
<br>
&gt; Question 2: Is it possible to get this &quot;compilation&quot; of the =
large XSD schema done, and then serialize the resulting java object to a fi=
le, and reload this pre-compiled thing so as not to face this compiling ove=
rhead at startup time?<br>
<br>
I find this is a good idea.<br>
<br>
Although currently not implemented, within Xerces, I think we could<br>
have an api to serialize and deserialize org.apache.xerces.xs.XSModel,<br>
or even javax.xml.validation.Schema to a file (using java&#39;s<br>
java.io.Serializable mechanism). I think, currently<br>
org.apache.xerces.xs.XSModel (and many of its dependencies) etc does<br>
not implement the interface java.io.Serializable, which we need to do<br>
so, if we need to implement this feature.<br>
<br>
We also would need to make performance measurements, if in case, we&#39;d<b=
r>
implement XSD schema serialization feature.<br>
<br>
If possibly, you could create a jira issue (new feature) for this<br>
requirement, that can make this requirement easy to track.<br>
<br>
<br>
-- <br>
Regards,<br>
Mukul Gandhi<br>
</blockquote></div>

--000000000000f1e44c05de5afe1c--