profile
viewpoint
If you are wondering where the data of this site comes from, please visit https://api.github.com/users/lemire/events. GitMemory does not store any data, but only uses NGINX to cache data for a period of time. The idea behind GitMemory is simply to give users a better reading experience.
Daniel Lemire lemire Université du Québec (TELUQ) Montreal, Canada http://lemire.me/en/ Daniel Lemire is a computer science professor. His research is on software performance in data engineering.

google/highwayhash 1180

Fast strong hash functions: SipHash/HighwayHash

bits-and-blooms/bitset 730

Go package implementing bitsets

fastfloat/fast_float 693

Fast and exact implementation of the C++ from_chars functions for float and double types: 4x faster than strtod

FastFilter/xorfilter 499

Go library implementing xor filters

FastFilter/xor_singleheader 197

Header-only Xor Filter library

FastFilter/fastfilter_java 169

Fast Approximate Membership Filters (Java)

FastFilter/fastfilter_cpp 145

Fast Approximate Membership Filters (C++)

eddelbuettel/rcppsimdjson 91

Rcpp Bindings for the 'simdjson' Header Library

geofflangdale/simdcsv 89

A fast SIMD parser for CSV files

push eventlemire/talks

Daniel Lemire

commit sha 94f6fbaa528420ef996795cbb26298daf99b91f8

Saving.

view details

push time in 8 hours

push eventRoaringBitmap/RoaringBitmap

Daniel Lemire

commit sha 6fc9e742f2bd7849e9fef96e999a2a8e59598e35

[Gradle Release Plugin] - new version commit: '0.9.22-SNAPSHOT'.

view details

push time in 14 hours

pull request commentRoaringBitmap/RoaringBitmap

Rangebitmap serde bug

@richardstartin Yes. I'll do the release right now.

For your convenience, I can give you my credentials but I would prefer not do it in clear text because then an adversary could hijack the project. Can we find a secured way to do so? Are you on telegram?

richardstartin

comment created time in 14 hours

created tagRoaringBitmap/RoaringBitmap

tag0.9.21

A better compressed bitset in Java

created time in 14 hours

push eventRoaringBitmap/RoaringBitmap

Daniel Lemire

commit sha dcb8f19722571fe91f419bb81166d959c54bd2de

[Gradle Release Plugin] - pre tag commit: '0.9.21'.

view details

push time in 14 hours

pull request commentRoaringBitmap/RoaringBitmap

Rangebitmap serde bug

Let me guess... another release in the near future? :-)

richardstartin

comment created time in 15 hours

pull request commentbits-and-blooms/bloom

Update bloom.go

Thanks.

moredure

comment created time in 20 hours

pull request commentRoaringBitmap/RoaringBitmap

fix boundary bugs, add test for floating point use case

The release is done.

richardstartin

comment created time in a day

pull request commentRoaringBitmap/RoaringBitmap

fix boundary bugs, add test for floating point use case

On the plus side, we no longer have a shitty process that takes 30 minutes.

richardstartin

comment created time in a day

push eventRoaringBitmap/RoaringBitmap

Daniel Lemire

commit sha 0c4b8968d60669191194a9654b1718571efcb654

[Gradle Release Plugin] - new version commit: '0.9.21-SNAPSHOT'.

view details

push time in a day

created tagRoaringBitmap/RoaringBitmap

tag0.9.20

A better compressed bitset in Java

created time in a day

push eventRoaringBitmap/RoaringBitmap

Daniel Lemire

commit sha 100024133c8b6b1f8dc26eced7bb5f5ec0e66535

[Gradle Release Plugin] - pre tag commit: '0.9.20'.

view details

push time in a day

pull request commentRoaringBitmap/RoaringBitmap

fix boundary bugs, add test for floating point use case

@richardstartin No burden at all.

Note that I do not create releases eagerly by design. So while it is a bit annoying, I want you to really ask for a release. Eager releases can create more trouble.

(I would also gladly give you the credentials if you want the ability to initiate the release, though I do not mind doing it.)

richardstartin

comment created time in a day

push eventsimdjson/simdjson

Daniel Lemire

commit sha d56fdc3d04e850de2afbd0e1bac39ca9115fd3c5

[no ci] deleting an irrelevant comment.

view details

push time in a day

push eventsimdutf/simdutf

Daniel Lemire

commit sha 844834508f5647b469792cce898400dc794fed69

Update README.md

view details

push time in 2 days

pull request commentRoaringBitmap/RoaringBitmap

fix boundary bugs, add test for floating point use case

Rerunning CI tests.

richardstartin

comment created time in 2 days

issue commentfastfloat/fast_float

Deviation from strtod

Thanks. Please keep reporting issues.

mrks

comment created time in 3 days

issue commentfastfloat/fast_float

Deviation from strtod

Swift: Screen Shot 2021-09-21 at 11 59 09 AM

mrks

comment created time in 3 days

issue commentfastfloat/fast_float

Deviation from strtod

I have removed any allusion to strtod in the README regarding the format specification, but note that I was paraphrasing the standard. It is strtod with some some changes (such as the leading +). That is what we try to follow. If we deviate from the standard, please submit an issue.

Screen Shot 2021-09-21 at 11 51 32 AM

mrks

comment created time in 3 days

pull request commentfastfloat/fast_float

constexpr fixes

I do not think that the C++17 specification requires from_chars to be constexpr.

mumbleskates

comment created time in 3 days

push eventfastfloat/fast_float

Daniel Lemire

commit sha 052975dd5f8166d0f9e4a215fa75a349d5985b91

Update README.md

view details

push time in 3 days

issue closedfastfloat/fast_float

Deviation from strtod

The readme states

It parses floating-point numbers expecting a locale-independent format equivalent to what is used by std::strtod in the default ("C") locale.

I noticed some difference between strtod and fast_float:

  • fast_float does not ignore leading whitespace as defined for strtod.
  • fast_float does not accept a leading 0x or +
  • For 1e999, strtod sets errno = ERANGE. fast_float returns inf and std::errc{} (i.e., no error). The other two libraries that I checked (absl::from_chars and boost::lexical_cast) also return an error.

Maybe the documentation could be extended to reflect the differences between strtod and fast_float?

closed time in 3 days

mrks

issue commentfastfloat/fast_float

Deviation from strtod

fast_float does not ignore leading whitespace as defined for strtod.

That's a per the from_char specification found, e.g., in the Microsoft documentation: When parsing chars, whitespace isn't ignored. Unlike strtod(), for example, the buffer must start with a valid numeric representation.

fast_float does not accept a leading 0x or +

The + sign is forbidden: the plus sign is not recognized outside of the exponent (only the minus sign is permitted at the beginning).

We do not support hexadecimal integers or hexadecimal floats at this time. Pull requests are invited.

For 1e999, strtod sets errno = ERANGE. fast_float returns inf and std::errc{} (i.e., no error)

Right. This is deliberate on my part. My view is that 1e999 is best represented as (positive) infinity.

Node (JavaScript):

Screen Shot 2021-09-21 at 11 24 34 AM

Python:

Screen Shot 2021-09-21 at 11 26 13 AM

I am open to changing this behaviour, but I'd like an open debate about it. Please open a specific issue if you think that the current behaviour should be changed.

I will close this issue and edit the README to address your concerns.

Thanks.

mrks

comment created time in 3 days

pull request commentfastfloat/fast_float

constexpr fixes

i am not completely convinced at the moment that there's a strong use case for constexpr parsing.

I also wonder. It is not often that from_char is called on a compile-time constant string.

mumbleskates

comment created time in 3 days

pull request commentbits-and-blooms/bloom

Update bloom.go

Yes, do an fmt please.

moredure

comment created time in 3 days

pull request commentfastfloat/fast_float

Don't use _BitScanReverse64 in constexpr context

@jrahlf It is all good. We want to cover GCC, clang and VS, if possible.

I will eagerly merge it, but we really want to have full tests because breaking people's code is not cool.

ryhor-spivak

comment created time in 3 days

issue closedlemire/JavaFastPFOR

Decompress using pyfastpfor

Can one use pyfastpfor to decompress the arrays compressed using JavaFastPFOR? Specifically looking for the following compression schemes:

val codec = SkippableIntegratedComposition(
            IntegratedBinaryPacking(),
            IntegratedVariableByte()
        )

and

 val codec = SkippableComposition(
            BinaryPacking(),
            VariableByte()
        )

closed time in 3 days

swapnilnawale

issue commentlemire/JavaFastPFOR

Decompress using pyfastpfor

Not to my knowledge.

If you need this feature, it should be possible to build it up, but it is possible that you might need to do some software work and add some additional tests.

Pull request invited!

swapnilnawale

comment created time in 3 days

push eventRoaringBitmap/RoaringBitmap

Daniel Lemire

commit sha 68cf6e2bf78e95957041d6f0efe464e7d5a30667

[Gradle Release Plugin] - new version commit: '0.9.20-SNAPSHOT'.

view details

push time in 4 days