Still has 1.1 declare-every-byte madness, then assemble, or partially assemble, a unified field elsewhere. But not in the same struct, or it will mess with the unsafe cast to make it operate fast enough for production.
Makes a mess of most of the straightforward options for converting the code. The coder also lacked the particular skill of faking a field in the struct by way of accessors. (Yea, my code uses such masking accessors, and I think it's a lot closer to ready for either .NET > 1.1 or porting to another language.)
The original code only had to deal with a hundred million records or so, and we've encountered some difficulties with it as the data has grown to approach two billion records. |
