Re: [sqlite] Conflict between snapshots and checkpoints

2019-09-29 Thread Gwendal Roué
Yes, this is exactly as you say. Isolation is very important, and I carefully make sure that it is guaranteed to users who need it. This has been working well for a few years now, GRDB.swift is very robust. My recent focus was on SQLite snapshots, and you and other members of this mailing list

Re: [sqlite] Conflict between snapshots and checkpoints

2019-09-29 Thread Richard Damon
On 9/29/19 11:40 AM, Gwendal Roué wrote: > Thank you very much Keith. > > Apologies for my imprecise vocabulary, and the use of the same "snapshot" > word with different meanings. > > I have used the term "snapshot isolation" as used in > https://www.sqlite.org/isolation.html; But I'll remember

Re: [sqlite] Conflict between snapshots and checkpoints

2019-09-29 Thread Gwendal Roué
Thank you very much Keith. Apologies for my imprecise vocabulary, and the use of the same "snapshot" word with different meanings. I have used the term "snapshot isolation" as used in https://www.sqlite.org/isolation.html; But I'll remember about REPEATABLE-READ isolation. I also thank you very

Re: [sqlite] Conflict between snapshots and checkpoints

2019-09-29 Thread Keith Medcalf
On Sunday, 29 September, 2019 01:28, Gwendal Roué wrote: >But now I fail to understand the indented use case of sqlite3 snapshots.. >Why allow to reuse snapshots with several calls to open()? Why do they >exist at all, since we can already profit from snapshot isolation with >one transaction

Re: [sqlite] Conflict between snapshots and checkpoints

2019-09-29 Thread Richard Damon
On 9/29/19 3:27 AM, Gwendal Roué wrote: > Thank you, Richard > > But now I fail to understand the indented use case of sqlite3 snapshots.. > Why allow to reuse snapshots with several calls to open()? Why do they > exist at all, since we can already profit from snapshot isolation with one >

[sqlite] Segfault in fts3IncrmergeLoad in v3.29

2019-09-29 Thread Conrad Irwin
Hi SQLite, I'd like to report a bug I've encountered using SQLite v3.29, where it reliably segfaults when trying to insert a row into a fts3 table in a database created with a previous version of SQLite. Please let me know if I can provide more information than what's below, as I'm keen to

Re: [sqlite] Segfault in fts3IncrmergeLoad in v3.29

2019-09-29 Thread Conrad Irwin
I've done some further digging here, and using git bisect, it seems like the problem was this change: https://sqlite.org/src/info/c736c40aab071a69. I can happily send a patch that fixes this for me, but I would like help producing a minimal test case to add to the test suite. In the database I

[sqlite] PRAGMA table_info fails with "no such tokenizer"

2019-09-29 Thread Anatoli Babenia
Python 3.7.4 (default, Jul 9 2019, 16:32:37) [GCC 9.1.1 20190503 (Red Hat 9.1.1-1)] on linux Type "help", "copyright", "credits" or "license" for more information. >>> import sqlite3 >>> conn = sqlite3.connect("/home/anatoli/.cache/tracker/meta.db") >>> cursor = conn.cursor() >>>

Re: [sqlite] Conflict between snapshots and checkpoints

2019-09-29 Thread Gwendal Roué
Thank you, Richard But now I fail to understand the indented use case of sqlite3 snapshots.. Why allow to reuse snapshots with several calls to open()? Why do they exist at all, since we can already profit from snapshot isolation with one transaction (at the cost of keeping it open)? For