lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Beady Geraghty" <>
Subject Re: file format of index
Date Fri, 14 Jul 2006 04:00:35 GMT
I think that I may be misreading the documentation.
I didn't see the description of the Long and Int type under the "Primitive
Types" section, while reading about the description of Byte, UInt32, Uint64,
VInt. So, for some reason I thought that Long and Int are byte
order sensitive.

Upon re-reading the document,  I see that
"All other data types are defined as sequences of bytes, so file formats are
byte-order independent. "    I think that I should be fine.

Sorry for posting before reading more carefully.

On 7/13/06, Beady Geraghty <> wrote:
> As I understand from earlier answers to my question that
> one can create an index on machine A,
> and use  it   (search and merge with other indices) on Machine B.
> I was reading the file format today.
> The index has Byte UInt32 UInt64 in most places, making it byte order
> indepdent in these places.
> However  a few spots has long and int.
> For example in the Compound Files section, there is a
> DataOffset-->Long
> and in term Vectors, there is a
> TVXVersion --> Int
> Is this an oversight in the documentation or is the document correct, and
> does this indicate that there will be problem using an index on a big
> Endian
> machine which is created on a small Endian machine.
> If both machines have the same Endian maybe the usage of Index created
> elsewhere is still fine ?
> I am starting to create quite a bit of code with the assumption that
> porting
> and merging index is okay anywhere.    I would appreciate some more
> input as to whether these fields mattered or not and whether the
> documentation
> is correct.
> Thank you in advance.

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message