Леонид Юрьев (Leonid Yuriev)

I suggest migrating to [MDBX](https://github.com/erthink/libmdbx), where this and other errors are fixed.

> oah, it already says "Included in the Awesome C list of open source C software." > > That's one way of positive thinking It it one more "awesome" list,...

@vlsi, no. Seems t1ha is too hardly for pure Java, because internally it uses extended multiplication of unsigned int64 with 128 bit result. e.g. https://docs.microsoft.com/en-us/cpp/intrinsics/umul128

1. t1ha needs 128-bit result of multiplication, therefore `imulq` is unsuitable, but `mulq` required. 2. Java don't provides an unsigned integer types or any kind of wide multiplication, therefore at...

> There's Math.multiplyHigh(long, long) since Java9, however it is not yet optimized by JIT compiler. Current t1ha headliner is t1ha1_le(), and it requires 64x64-to-128 _unsigned_ multiplication. For future versions, e.g....

> Well, Critical JNI seems to be able... Yes, **Critical Natives** give a chanse for `t1ha1()`. [JDK-7013347](https://bugs.openjdk.java.net/browse/JDK-7013347) ``` JNIEXPORT jlong JNICALL JavaCritical_com_package_t1ha_t1ha1_le(jint length, jbyte* buf, jlong seed) { return t1ha1_le(buf,...

@oddjobz, this issue may be interested for you in `pymamba` context: - `libfpta` based on [libmdbx](https://github.com/ReOpen/libmdbx), which is a fork of LMDB. - `libfpta` provides secondary indexes. - `libfpta` provides...

These restrictions come from the original DNA of LMDB. On the other hand, such restrictions caused by understanding specific of LMDB/mdbx, with a bit of common sense :)

Not 'limited', but `libfpta` (and LMDB also) may be not the best choice if .

Actual 'the hard' limits are comes from [libfptu](https://github.com/leo-yuriev/libfptu), exactly from [here](https://github.com/leo-yuriev/libfptu/blob/master/fast_positive/tuples.h#L202-L255): ``` fptu_max_tuple_bytes = 262140 // maximum total size of the serialized representation of the tuple fptu_max_cols = 1022 //...