adsharma

10 ms·

121.
by adsharma 10mo ago
It's interesting to see people use the term "GQL" to refer to GraphQL. https://www.gqlstandards.org/ is an ISO standard. The Graph Database people don't love search engine results when they're looki
122.
by adsharma 10mo ago
For every person trying to move an old code base from COBOL to Java to remove tech debt, there are an equal number of people who want rewrite a working C++ code base in Rust/Go/Zig. Leaders who know that it's a people problem
123.
by adsharma 11mo ago
We need a new word for a universal transpiler. Something that can transpile to 7 or more languages. Poly-transpiler? It will also trigger more people.
124.
by adsharma 11mo ago
Py2many would fail to compile dynamic code whereas spy runs it through the interpreter. Otherwise the internal structure looks similar. Py2many has been around for 10+ years under previous names and has significant test coverage.
125.
by adsharma 11mo ago
See my other comment in the thread. I would argue that anything that uses arcane dynamic stuff in python should be renamed to .dpy and the vast majority of the commonly used constructs retain .py The issue in HN threads like this is that ev
126.
by adsharma 11mo ago
The position I take is that such obscure code in the guts of a popular package could be slowing down large amounts of deployed code elsewhere. If such code must exist, it should be marked as special (like how Cython does it). Beyond adding
127.
by adsharma 11mo ago
Where is it? Would love to compare the approach to py2many.
128.
by adsharma 11mo ago
++spy ethos and ideas But why limit to an interpreter? Translate to other excellent compiled languages and benefit from the optimization work there. Giving up on C-API and the dynamic parts of python that 1% of the people use is a good trad
129.
by adsharma 11mo ago
https://www.epcc.ed.ac.uk/whats-happening/articles/exploring...
130.
by adsharma 11mo ago
One of the problems with Ceph is that it doesn't operate at the highest possible throughput or the lowest possible latency point. DAOS seemed promising a couple of years ago. But in terms of popularity it seems to be stuck. No Ubuntu p
131.
by adsharma 11mo ago
Ceph is under LGPL.Cost doesn't seem to be a barrier. Supports k8s through CSI and has observability and documentation. You can probably hire people to maintain it. Was it the ramp-up cost or expertise?
132.
by adsharma 1y ago
https://ladybugdb.com/ is a fork of Kuzu.
133.
by adsharma 1y ago
Morsel driven parallelism is working great in DuckDB, KuzuDB and now Ladybug (fork of Kuzu after archival).
134.
by adsharma 1y ago
I saw one fleeting reference to a CNI in the article. Anyone familiar with the space will tell you this is the biggest blocker in production. You will have to pay for an "enterprise" CNI to make it work.
135.
by adsharma 1y ago
Love the DuckLake spec. Now there is something similar for Graphs. https://adsharma.github.io/graph-archiving/ https://adsharma.github.io/beating-the-CAP-theorem-for-graph... The catalog is based on th
136.
by adsharma 1y ago
This URL works for now: https://kuzudb.github.io/docs/developer-guide/database-inter...
137.
by adsharma 1y ago
The code to implement this was open sourced today: https://github.com/LadybugDB/ladybug/pull/3
138.
by adsharma 1y ago
DuckDB is scalable (can handle TPC-H 1TB) and open source, but doesn't support graphs natively. It supports some graph queries on a SQL native columnar storage. With the proposed solution, you'll be able to query larger graphs on
139.
by adsharma 1y ago
That's right - it was a fun 2 out of 3 analogy. The real question being raised in the blog post is - should the next generation graph databases pursue a local-only embedded strategy or build on top of object storage like many non-graph
140.
by adsharma 1y ago
One thing you might want to watch out for is that the storage format on disk is not stabilized. Last few releases, you couldn't open a file written by a previous version of kuzu. You had to constantly export/import as new versions
141.
by adsharma 1y ago
PGQ requires you to write using SQL and read using a graph query language. GQL is a standalone language that supports reads/writes. But much of the community is still using cypher. More on this here: https://adsharma.github.
142.
by adsharma 1y ago
Link to kuzudb internals: https://kuzudb.com/docs/developer-guide/database-internal/
143.
by adsharma 1y ago
A more interesting path is to keep dbscan_inner in pure python with type annotations and then use py2many --mojo=1 dbscan_inner.py to translate.
144.
by adsharma 1y ago
Recent: 3.6 bits per param https://arxiv.org/abs/2505.24832
145.
by adsharma 1y ago
2 bits out of FP8 would be 25% 2 bits out of FP16 would be 12.5% I've seen recent work that claimed 70% of the params are used for memorization.
146.
by adsharma 1y ago
I wish the authors calculated a plot of model size (number of params) vs number of triples it can hold before the memory collapse happens. It's hard to map the frequency of knowledge injection to a real world understanding of "how
147.
by adsharma 1y ago
The weakness of these systems is not the query language or the planner. It's the lack of a fully developed storage engine that avoids vendor lock-in. Apache GraphAr (incubating) is a step in this direction. But it's an import&#x2F
148.
by adsharma 1y ago
Those of you wondering about how to use z3, please consider coding in static python (not z3py) and then transpile to smt2. You'll be able to solve bigger problems with familiar/legible syntax and even use the code in production.
149.
by adsharma 1y ago
Python -> Mojo -> MLIR ... <target hardware> Yes, you can write mojo with python syntax and transpile. You'd end up with something similar to Julia's 1.5 language problem. Since the mojo language is not fully s
150.
by adsharma 1y ago
Use whatever dhh recommended last year! I like parts of omakub, but impossible to use alacritty on Windows. Paste is broken. Known WONTFIX. I would use a tiling window manager, but not ready to give up on the mouse. Title bars are sometimes

More ›