v2.2.0

more options, more explanations

v2.1.0

NEW ALGORYTHEMS, NEW COMANDLINE OPTIONS, CORRECTLY HANDLE THE DELETE BYTE IN DBFS

v2.0.2

you can now import this library directly with

import Coshp from 'https://unpkg.com/coshp@latest/coshp.js';

v2.0.1

Accidentally forgot to include the dist folder in the package published to npm, you need that for the projections.

v2.0.0

Version 2.0.0 corrects enough bugs in how .qix files are generated and handled that if you happened to actually be using this in production (you should not) I'd consider it a breaking change. The upside is that it's totally compatible with mapsever now. Changes include

• correctly calculating .qix sibling node offsets, the .qix documentation had a bug implying they were calculated from the end of the mandatory part of the node, they are actually calculated from the end of the node.
• zero indexing all ids in the .qix files, shp files are technically 1 indexed but since record ids are usually inferred from position the actual id values are rarely dealt with.
• did you know there is a random byte between the header and the rest of the file in .dbf file? Yeah neither did I but it's now included in the .dbf files created when reordering your files with this tool.
• fixed a bug where we were accidentally not grabbing the last record of any range we queried
• .qix files are now loaded incrementally in user specified blocks defaulting to 4kb.
• downloading rows from the remote .shp and .dbf file is now parallelized up to a limit.

v1.1.0

This changes the algorithm for generating .qix files to a packed Hilbert R-tree based on the implementation in flatbush. In some limited testing I've a similar number of ranges of rows being requested but FAR fewer rows need to be filtered out, (~1500 vs ~30).