My vote would be to replace spaces w/ underscores.

Could probably be solved with a ~6 line subclass of FieldMutatingUpdateProcessor

Chris Hostetter (Unused), I tried this:

public class NormalizeFieldNameUpdateProcessorFactory extends FieldMutatingUpdateProcessorFactory {
  @Override
  public UpdateRequestProcessor getInstance(SolrQueryRequest req, SolrQueryResponse rsp, UpdateRequestProcessor next) {
    return new FieldMutatingUpdateProcessor(getSelector(), next) {
      @Override
      protected SolrInputField mutate(SolrInputField src) {
        src.setName(src.getName().replace(' ', '_'));
        return src;
      }
   };
  }
}

And got this error:

<lst name="error"><str name="msg">mutate returned field with different name: field with spaces =&gt; field_with_spaces</str><str name="trace">org.apache.solr.common.SolrException: mutate returned field with different name: field with spaces =&gt; field_with_spaces...

Are there problems that would result when changing the name of a field in FieldMutatingUpdateProcessor?

Are there problems that would result when changing the name of a field in FieldMutatingUpdateProcessor?

i suspect i put that in as a sanity check to protect the the surface area of the API – i don't know if relaxing that will cause problems, or if it's just something that's there because the ramifications of allowing it aren't really well tested in the rest of the FieldMutating code paths.

in particular: what does it mean? should the old field name be removed? should the corisponding field:value pair be rmeoved, but other instances of that field:value2 be left in (ie: what if the mutator renames one instance of the field but not another?)

easiest thing would probably be to implement field renaming it as a complete one-off special UpdateProcessor w/o using hte FieldMutating framework (ie: no config, just something barebones for use in schemaless that can maybe later be re-parented in the class hierarchy to support more config options)

A new URP called FieldNameMutatingUpdateProcessorFactory
example configuration to replace spaces with underscores

    <processor class="solr.FieldNameMutatingUpdateProcessorFactory">
      <str name="pattern">\s</str>
      <str name="replacement">_</str>
    </processor>

no test cases yet

Noble Paul] - looks good! The pattern should be expanded to include all the funky/problematic/illegal characters before committing, but in general +1.

The only catch is , if there are multiple patterns to match you need multiple <processor> tags . I hope it is OK

+1

with tests
and now replaces all non word chars except hyphen -

Commit 1651587 from Noble Paul in branch 'dev/trunk'
[ https://svn.apache.org/r1651587 ]

SOLR-6937 In schemaless mode ,replace spaces and special characters with underscore

Commit 1651588 from Noble Paul in branch 'dev/branches/branch_5x'
[ https://svn.apache.org/r1651588 ]

SOLR-6937 In schemaless mode ,replace spaces and special characters with underscore

Commit 1651589 from Noble Paul in branch 'dev/branches/lucene_solr_5_0'
[ https://svn.apache.org/r1651589 ]

SOLR-6937 In schemaless mode ,replace spaces and special characters with underscore

Commit 1651646 from Noble Paul in branch 'dev/trunk'
[ https://svn.apache.org/r1651646 ]

SOLR-6937 don't replace periods

Commit 1651647 from Noble Paul in branch 'dev/branches/branch_5x'
[ https://svn.apache.org/r1651647 ]

SOLR-6937 don't replace periods

Commit 1651648 from Noble Paul in branch 'dev/branches/lucene_solr_5_0'
[ https://svn.apache.org/r1651648 ]

SOLR-6937 don't replace periods

Commit 1652651 from Noble Paul in branch 'dev/branches/lucene_solr_5_0'
[ https://svn.apache.org/r1652651 ]

reverting SOLR-6937

Commit 1652967 from Noble Paul in branch 'dev/branches/lucene_solr_5_0'
[ https://svn.apache.org/r1652967 ]

SOLR-6937 In schemaless mode ,replace spaces and special characters with underscore, This impacts usability so ,it is voted to be in 5.0

Bulk close after 5.0 release.