On Jan 28, 2008 4:47 PM, Yousef Ourabi <[EMAIL PROTECTED]> wrote: > Yonik could you elaborate a bit -- I'm not following .. > > 1) The RubyResponseWriter should never leave a hash value blank in a (k,v) > pair, it should use Nil -- this is handled in the patch. What is missing? > This, in my opinion is separate from how that value got empty/nil in the > index. > > 2) Which leads to Schema validation which the second patch handles. > > What part of the picture am I not seeing?
fix_ruby_output.patch (this was a patch by someone else), only fixes ruby output and is hence incomplete and implemented in the wrong place. zero_length_int.patch (my quick hack), was for changing a zero length string to a 0 for integer (one possible interpretation for this weird case)...but it's still for the output side, not validating input at index time. So we should fix both #1 and #2, but hopefully you agree it's not the highest priority since it doesn't block anyone from fixing their client indexing code (which would still break after we fixed #2). -Yonik
