Add Location of a Problem if @line Attribute is Absent

[volodya-lombrozo]

We have found, that if @line attribute isn't set, we can't fide the place where the issue occurred, for example:

[[j$Application application-duality CRITICAL]:0 The anonymous object has duality in the application, which is prohibited,

Instead of 0 we have to put something meaningful.

There are several suggestions how to solve this issue:

#320 (comment)

#320 (comment)

[volodya-lombrozo]

@yegor256 What do you think?

[yegor256]

@volodya-lombrozo I'm not sure it's solvable, but we can try

[h1alexbel]

@yegor256 I'm in

[yegor256]

@h1alexbel sure, please, help!

[h1alexbel]

@yegor256 @volodya-lombrozo how about we introduce new <context/> node inside the <defect/>? For instance, if there are no lines, application-duality will produce such a defect:

<defect line="0" severity="critical">The anonymous object has duality in the application, which is prohibited
  <context>
    <o base="man">
      <o as="name" base="string"/>
      <o base="number"/>
    </o>
  </context>
</defect>

In each XSL, where it is possible, we are able to copy the problematic XML like this <xsl:copy-of select="."/>.

The users of EO don't need this <context/> to be printed, since EO-to-Java compiler places lines, while other users, who generate XMIR by themselves (like JEO) may find it useful using lints library:

final Defect defect = ...;
final XML context = defect.context();

WDYT?

[volodya-lombrozo]

@h1alexbel I like this idea. It might slightly help in my particular case. (Of course, getting an exact line would be perfect) However, there is one potential issue I’m concerned about: the context size. If it's small enough, then it's fine. But if it is large, it might create some performance issues. What do you think?

[h1alexbel]

@volodya-lombrozo yes, <context/> can be bigger than reasonable size, but depends on specific lint scope. To resolve potential performance issues, we can try to convert <context/> to string, and then print only first N chars of the context. WDYT?

[h1alexbel]

@volodya-lombrozo About lines I agree, but on the other hand, it can add some processing overhead, since we need to annotate original XMIR with lines, resolve them, and then pass annotated XMIR to lints. @yegor256 Should we try to go for it?