Project

Profile

Help

Forums » Help »

Design of the Namepool in Saxon-B

Added by Anonymous about 18 years ago

Legacy ID: #3915931 Legacy Poster: oneil (ond1)

Hi, I would like to know why the Namepool data structure was designed in this way (in Saxon-b)? What was the reason for hashing only on the local name for the Namepool data structure? Where there any construction or running advantages in the design of the data structure? If the hash-function was on the URI+local name, having one large table would there be any pitfalls? regards, O'Neil


Replies (1)

RE: Design of the Namepool in Saxon-B - Added by Anonymous about 18 years ago

Legacy ID: #3915976 Legacy Poster: Michael Kay (mhkay)

There are many pragmatic decisions in the design of the NamePool: some of these decisions were fairly arbitrary and experimental at the time, but I haven't found a need to change them (that is, there haven't been any identified performance problems that would be solved by a change to the design). The rationale for hashing on the local name is that it's relatively unusual to have the same local name in several namespaces; hashing on a long string like a URI is expensive, and would do very little to reduce the number of collisions in most common use cases.

    (1-1/1)

    Please register to reply