epubcheck with error while using tex4ebook and \nocite bibtexing

Question

While running biblatex with \nocite commands, I got the following error message(s) when running epubcheck after tex4ebook.

ERROR(RSC-005): Testfile.epub/OEBPS/Testfileli3.html(19,88): Validierungsfehler: Der Wert des Attributs "id" ist ungültig; es muss ein XML-Name ohne Doppelpunkte sein

Translated:
validation error: The value of attribute "id" is invalid; is has to be a XML-name without colon

The MWE would be the following, but it only reproduced the error 1 time. In my sourcefile.tex I got 175 \nocite commands and got 173 identical error messages with epubcheck.

\documentclass[11pt,a4paper]{report}
\usepackage{ebgaramond-maths}
\usepackage[T1]{fontenc}
\usepackage[utf8]{inputenc}
\usepackage[ngerman]{babel}
\usepackage{tex4ebook}
\usepackage[sortlocale=auto,bibstyle=authoryear,citestyle=authortitle-ticomp]{biblatex}
\addbibresource{biblatex-examples.bib}

\begin{document}

\nocite{sigfridsson}
\nocite{westfahl:space}
\nocite{set}
\nocite{stdmodel}
\nocite{aksin}
\nocite{bertram}

\printbibliography

\end{document}

Is there a fix for it? --> Yes, see Answer 1. It worked reducing the epubcheck error number from 123 to 20.

Update 1

I now have 20 errors left with epubcheck, which seem all related to the \pageref{somelabel} command. epubcheck gives ERROR(RSC-012) Fragmentbezeichner ist nicht angegeben and points to the following example in the html files (out of the 20):

(wie vorangehend, bei Andrej
Sacharow<a id='dx17-18004'></a> ab Seite <a href='#x11-12001r9'>249<!--  tex4ht:ref: AndrejSacharow   --></a>, gesehen und angedeutet)

and points to the page number "249". But this \pageref is located in ordinary body of text, without embedding in other command and env.

Up to now it was not possible to reproduce the epubcheck error in a MWE.

Answer 1

This issue should be fixed by make4ht DOM filters, but I found that this particular filter is broken. I've fixed that in make4ht sources, bur if you don't want to update it, you can use this build file instead:

local domfilter = require "make4ht-domfilter"

local allowed_chars = {
  ["-"] = true,
  ["."] = true
}
local function fix_colons(id)
  -- match every non alphanum character
  return id:gsub("[%W]", function(s)
    -- some characters are allowed, we don't need to replace them
    if allowed_chars[s] then return s end
    -- in other cases, replace with underscore
    return "_"
  end)
end

local function id_colons(obj)
  -- replace : characters in links and ids with unserscores
  obj:traverse_elements(function(el) 
    local name = string.lower(obj:get_element_name(el))
    if name == "a" then
      local href = el:get_attribute("href")
      -- don't replace colons in external links
      if href and not href:match("[a-z]%://") then
        local base, id = href:match("(.*)%#(.*)")
        if base and id then
          id = fix_colons(id)
          el:set_attribute("href", base .. "#" .. id)
        end
      end
    end
    local id  = el:get_attribute("id")
    if id then
      el:set_attribute("id", fix_colons(id))
    end
  end)
  return obj
end


local process = domfilter {id_colons}

Make:match("html$", process)

Compile using:

I also found another issue, and that is a wrong formatting of DOI numbers. It can be fixed using this configuration file:

\Preamble{xhtml}
\def\nolinkurl#1{#1}
\begin{document}
\EndPreamble

Compile using:

$ tex4ebook -e build.lua -c config.cfg filename.tex

This is the result: