biblatex authoryear-comp and name aliases

Question

Using the authoryear-comp style I often cite different authors having the same last name. The biblatex package options

backend=biber,style=authoryear-comp,firstinits=true,uniquename=init

eliminate ambiguities by adding author initials to these citations. Problem is that I also cite some prolific authors who don't consistently use their middle name or hyphenate their first and middle names.

For example in the document below I get the citation:

J.-P. Doe 2005; J. Doe 2006; J. P. Doe 2007; J. P. Doe and Jones 2008; J.-P. Doe and M. Smith 2009; M. Smith and Jones 2011; A. Smith and J. Smith 2011

but there is really only one Doe here. Name inconsistencies for Doe in the reference list are fine as they accurately reflect the bibliographic data. But I'd rather have the citation be:

Doe 2005, 2006, 2007; Doe and Jones 2008; Doe and M. Smith 2009; M. Smith and Jones 2011; A. Smith and J. Smith 2011

Is there an easy way to define author/editor name aliases? The shortauthor field offers one approach, but ideally I'd like to avoid ongoing edits to the bib file.

\documentclass{article}
\usepackage[american]{babel}
\usepackage{csquotes}
\usepackage[backend=biber,style=authoryear-comp,firstinits=true,uniquename=init,sorting=none]{biblatex}

\usepackage{filecontents}
\begin{filecontents}{\jobname.bib}
@Book{hyphen,
  author = {Doe, John-Paul},
  title = {A book by John-Paul Doe},
  publisher = {Publisher},
  year = {2005}}
@Book{nomiddle,
  author = {Doe, John},
  title = {Another book by John-Paul Doe},
  publisher = {Publisher},
  year = {2006}}
@Book{nohyphen,
  author = {Doe, John Paul},
  title = {Yet another book by John-Paul Doe},
  publisher = {Publisher},
  year = {2007}}
@Article{initials,
  author = {Doe, J. P. and Jones, J.},
  title = {An article coauthored by John-Paul Doe},
  journal = {Journal title},
  year = {2008}}
@Article{hypheninitials,
  author = {Doe, J.-P. and Smith, M.},
  title = {Another article coauthored by John-Paul Doe},
  journal = {Journal title},
  year = {2009}}
@Article{smithm,
  author = {Smith, Mary and Jones, Jane},
  title = {Article title},
  journal = {Journal title},
  year = {2011}}
@Article{smitha,
  author = {Smith, Anne and Smith, Joe},
  title = {Article title},
  journal = {Journal title},
  year = {2011}}
\end{filecontents}
\addbibresource{\jobname.bib}

\begin{document}
\cite{hyphen,nomiddle,nohyphen,initials,hypheninitials,smithm,smitha}
\printbibliography
\end{document}

Answer 1

Here is one approach which gives exactly the result you want. Put this in your biber.conf:

<?xml version="1.0" encoding="UTF-8"?>
<config>
  <sourcemap>
    <maps datatype="bibtex" map_overwrite="1">
      <map maptype="field">
        <map_pair map_source="AUTHOR"/>
        <also_set map_field="SHORTAUTHOR" map_origfieldval="1"/>
        <also_set map_field="SORTNAME" map_origfieldval="1"/>
      </map>
      <map maptype="field">
        <map_pair map_source="SHORTAUTHOR" map_match="Doe,\s*J(?:\.|ohn)(?:[ -]*)(?:P\.|Paul)*" map_replace="Doe, John Paul"/>
        <map_pair map_source="SORTNAME" map_match="Doe,\s*J(?:\.|ohn)(?:[ -]*)(?:P\.|Paul)*" map_replace="Doe, John Paul"/>
      </map>
    </maps>
  </sourcemap>
</config>

This follows the new Biber 0.9.8 semantics for the maps - the maps are processed sequentially, each depending on the results of the previous. Within a map, you can also use multiple items but items within a map which change things should not rely on the results of other other changes within the map element. Use map elements to do the sequencing. So, this first copies AUTHOR to both SHORTAUTHOR and SORTNAME and then, after that, edits these two fields to regularise some specific names. Biber DEV PDF doc will be updated soon.

It's a rather hairy regular expression but they are incredibly useful for faking semantic equivalence. This essentially renders the Does into the canonical "Doe, John Paul" format so that Biber knows that they are the same person. It's probably possible to design a regular expression to make this not so name specific but that will depend in the idiosyncracies of the name formats. For example, you could remove all "-' characters in names and also, since you are using firstinits, force all names to initials only.

EDIT: There will be a simpler syntax in Biber 0.9.8 like this:

<?xml version="1.0" encoding="UTF-8"?>
<config>
  <sourcemap>
    <maps datatype="bibtex" map_overwrite="1">
      <map>
        <map_step map_field_source="AUTHOR" map_match="Doe," map_final="1"/>
        <map_step map_field_set="SHORTAUTHOR" map_origfieldval="1"/>
        <map_step map_field_set="SORTNAME" map_origfieldval="1"/>
        <map_step map_field_source="SHORTAUTHOR" map_match="Doe,\s*J(?:\.|ohn)(?:[ -]*)(?:P\.|Paul)*" map_replace="Doe, John Paul"/>
        <map_step map_field_source="SORTNAME" map_match="Doe,\s*J(?:\.|ohn)(?:[ -]*)(?:P\.|Paul)*" map_replace="Doe, John Paul"/>
      </map>
    </maps>
  </sourcemap>
</config>

Answer 2

I pondered about something similar recently: If the same author has published books under different variants of his name as you've described, should they really be printed differently in the bibliography? For example, if you have a style with the dashed option where recurring names are replaced by a dash, should the variations be treated as completely different names and not be replaced? I chose to unify the names in the .bib file.

I think an important question is whether we're really talking about accuracy here or only pseudo-accuracy. For example, I've often seen that the cover of a book says "J.P. Doe" while the first page says "John Paul Doe".

Another question is whether you really run the risk of confusing authors when you go for consistency in the bibliography in this way. After all, the main goal of any bibliography is to ensure that your reader can easily find the source you're citing. If you have different entries for "J.P. Doe" and "John Paul Doe", don't you actually make things more difficult for your reader?

So if it's only a question of initials vs. full names and the like I would opt for actually standardizing the .bib file.