Closed progval closed 2 months ago
Yes, but the idea was that of putting them elsewhere rather than in the compression code
BTW is there any specific stat you had in mind?
On August 1, 2024 11:51:43 AM GMT+02:00, Val Lorentz @.***> wrote:
The Rust rewrite of permute removed a bunch of properties from
graph.stats
: https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fvigna%2Fwebgraph-big%2Fblob%2F51f1ebec826397a79c500e8c09e7ae7192015285%2Fsrc%2Fit%2Funimi%2Fdsi%2Fbig%2Fwebgraph%2FBVGraph.java%23L2490-L2520&data=05%7C02%7Cvigna%40di.unimi.it%7C6498c108cf344b7dc9a008dcb20f8d65%7C13b55eef70184674a3d7cc0db06d545c%7C0%7C0%7C638581027074200949%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C0%7C%7C%7C&sdata=9zds9ErOPuulHIpMhl4n6ZIDto1Hi3VGtCvkZrcJAC4%3D&reserved=0 -> https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fvigna%2Fwebgraph-rs%2Fblob%2Faf48d3b8f98b2378ebf50e2ea536947bd60c65d8%2Fsrc%2Fgraphs%2Fbvgraph%2Fcomp%2Fflags.rs%23L89-L170&data=05%7C02%7Cvigna%40di.unimi.it%7C6498c108cf344b7dc9a008dcb20f8d65%7C13b55eef70184674a3d7cc0db06d545c%7C0%7C0%7C638581027074357161%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C0%7C%7C%7C&sdata=o8zdzLq5iKYmdmqIUF8K%2BBJdbdgTiBRCqJFLfrcQI54%3D&reserved=0Are you planning to re-add them?
-- Reply to this email directly or view it on GitHub: https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fvigna%2Fwebgraph-rs%2Fissues%2F109&data=05%7C02%7Cvigna%40di.unimi.it%7C6498c108cf344b7dc9a008dcb20f8d65%7C13b55eef70184674a3d7cc0db06d545c%7C0%7C0%7C638581027074357161%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C0%7C%7C%7C&sdata=mQutdjFaQruB%2FnN3KNfKOAsXzsRXfDhmqWHQm0ZAYL8%3D&reserved=0 You are receiving this because you are subscribed to this thread.
Message ID: @.***>
compratio, bitspernode, and bitsperlink seem to be the only ones we used at any point at SWH
Oh wow. We don't do even do those? That's bad—I thought the distribution of gaps or such. That should be easy to fix.
I'll add them, compress ratio is computed based on what?
It is computed with respect to the information-theoretical lower bound. You can have a look at the Java code.
Done, I checked that on cnr2000 all the parameters match, the only difference is the number of decimal places; On Java is 3 decimals, in this impl is not truncated so ~15.
Should I also truncate it to 3 decimals?
I also added the length
field, which contains the length in bits of the compressed graph.
The Rust rewrite of permute removed a bunch of properties from
graph.stats
: https://github.com/vigna/webgraph-big/blob/51f1ebec826397a79c500e8c09e7ae7192015285/src/it/unimi/dsi/big/webgraph/BVGraph.java#L2490-L2520 -> https://github.com/vigna/webgraph-rs/blob/af48d3b8f98b2378ebf50e2ea536947bd60c65d8/src/graphs/bvgraph/comp/flags.rs#L89-L170Are you planning to re-add them?