[lmi] Empty paragraphs in HTML-MST

[Greg Chicares]

Motivation: MST files often contain blocks like this example:
  <p>{{SampleText}}</p>
which does the right thing if 'SampleText' is nonempty. However,
for certain products, 'SampleText' is deliberately empty, in
which case the example above expands to '<p></p>', printing an
empty paragraph that consumes vertical space [0]. We'd like to
suppress that effect.

One option requires defining a new variable 'SampleTextIsEmpty':
  {{#SampleTextIsEmpty}}
    <p>{{SampleText}}</p>
  {{/SampleTextIsEmpty}}
but that's verbose, and variables should not be multiplied
without necessity.

Before attempting to devise a simpler way, I'd like to ask
whether the observed behavior is necessarily correct. HTML4
suggests the desired behavior (no effect):

  https://www.w3.org/TR/REC-html40/struct/text.html#edef-P
| We discourage authors from using empty P elements.
| User agents should ignore empty P elements.

The HTML3 specification (upon which wxHTML is based IIRC)
is less clear:

  https://www.w3.org/MarkUp/html3/paras.html
| Do not use empty paragraphs to add white space around
| headings, lists or other elements. White space is added
| by the rendering software.

but would it make sense for wxHTML to ignore empty paragraph
elements, as a runtime option?

If not, then is the tidiest solution to add an extra step
equivalent to
  s/<p> *<\/p>//g
to lmi's MST processing?

Here's one particular example, in 'ill_reg_column_headings.mst'
(simplified):

{{#SinglePremium}}
<p>
<b>Ultimate Illustrated Crediting Rate:</b>
{{UltCreditingRateFootnote}}
{{/SinglePremium}}
</p>

Today, some products set <UltCreditingRateFootnote> to
  "[specimen text]"
and should print
  <p><b>Ultimate Illustrated Crediting Rate:</b>[specimen text]</p>
of course. But now a new product that meets the {{#SinglePremium}}
condition has no distinct "ultimate" rate, and shouldn't have this
footnote. If we just change <UltCreditingRateFootnote> to "", we get
  <p><b>Ultimate Illustrated Crediting Rate:</b></p>
which is of course wrong. I'd like to bring the (bold) name of the
term into <UltCreditingRateFootnote> in the '.policy' files:
  <UltCreditingRateFootnote>
    <strong>Ultimate Illustrated Crediting Rate:</strong>[description]
  </UltCreditingRateFootnote>
for the original product that has a non-empty description, but
  <UltCreditingRateFootnote/>
for the new product that has no "ultimate" rate (and shouldn't have
such a footnote at all--not even a blank space). Then the MST becomes:
  <p>{{UltCreditingRateFootnote}}</p>
where we'd want to remove any resulting empty "<p></p>". Alternatively,
the MST might specify merely:
  {{UltCreditingRateFootnote}}
which is terser, but moves markup out of the MST file (and it could be
debated whether that markup is "semantic" or "presentational").

I certainly don't want to introduce a new product-database field like
  {DB_HasUltCreditingRate
  ,DB_Topic_WhoEvenKnows
  ,"HasUltCreditingRate"
  ,"Print a footnote describing an ultimate rate: 0=no, 1=yes",}
inducing changes in dozens of source files, hundreds of generated
product files, and a thousand regression-test files; or try to
synthesize a new variable in 'pdf_command_wx.cpp' like
  add_variable
    ("HasUltCreditingRate"
    ,...analyze 'AnnGAIntRate_Current' to see whether it varies
     during some period that cannot actually be defined, which
     necessarily makes the footnote's conditional dynamic, but
     it has to be static, so...that idea is foredoomed
    );

The design inherited from XSL-FO has this conditional:
  {{#SinglePremium}}
which is set here:
  add_variable
    ("SinglePremium"
    ,  starts_with(policy_name, "Single")
    || starts_with(policy_name, "Modified")
    );
and of course that bletchery could be compounded in a way that
might happen to do the right thing for the new product that has
no "ultimate" rate--but that's like defining
  bool CarIsAmerican(char* marque)
    {return 'F' == marque[0];}
because 'F' is the first letter of "Ford". Oops--"Fiat":
    {return 'F' == marque[0] && 'i' != marque[1];}
Then we add another special case for "Ferrari". Years later,
someone points out that "Chevrolet" doesn't start with "F".
All these funky conditionals must die.

Thus, the product --> HTML --> MST chain is excessively complex
in multiple ways, but we can't eradicate the complexity until
we figure out a clean way to suppress empty <p>footnotes</p>.

---------

[0] "an empty paragraph...consumes vertical space"

Proof: insert the following in an '.mst' file, then use it to
print an illustration.

<p>Here follows one empty paragraph...</p>
<p></p>
<p>...and three empty paragraphs...</p>
<p></p> <p></p> <p></p>
<p>...ending here.</p>