PCDLoader: parse header without TextDecoder

[F-loat]

Related issue: -

Description

More efficient parsing of header information.

[Mugen87]

How does these numbers compare to the previous version of PCDLoader?

[F-loat]

@Mugen87 The ascii in the above figure represents the time taken by the previous version, and the binary represents the time taken by new version.

The parsing time of the old version will increase significantly as the pcd file becomes larger, while the new version keeps a relatively low parsing time.

[Mugen87]

Is the performance different that large because TextDecoder essentially parses the entire data and not just the header?

AFAIK, PCD headers do not have a fixed size so it is not possible to detect the header portion of the binary data. TextDecoder must decode everything which is of course slow, right?

In this case, the proposed change looks good

[F-loat]

Yes, for the reason you said TextDecoder can only parse the entire file, which is wasteful.