owner stringclasses 14
values | repo stringclasses 14
values | id int64 116k 4.61B | issue_number int32 1 181k | author stringlengths 1 39 | body stringlengths 1 262k | created_at timestamp[us]date 2000-06-06 02:40:44 2026-06-02 20:16:18 | updated_at timestamp[us]date 2000-06-06 02:40:44 2026-06-02 20:22:56 | reactions unknown | author_association stringclasses 5
values |
|---|---|---|---|---|---|---|---|---|---|
ClickHouse | ClickHouse | 593,406,198 | 9,461 | alexey-milovidov | > no member named 'contains'
It requires recent version of libc++. We have it in `contrib` directory (libcxx, libcxxabi).
I don't know why `contrib` version of libc++ is disabled on FreeBSD. Most likely just in case and nobody paid any attention to it yet. | 2020-03-02T13:38:19 | 2020-03-02T13:38:19 | {} | MEMBER |
ClickHouse | ClickHouse | 593,450,155 | 9,300 | alexey-milovidov | The overall impression is that it is made in "тяп-ляп" fashion.
Example:
| 2020-03-02T15:11:08 | 2020-03-02T15:11:08 | {} | MEMBER |
ClickHouse | ClickHouse | 593,719,541 | 9,472 | alexey-milovidov | It requires further investigation. | 2020-03-03T01:31:13 | 2020-03-03T01:31:13 | {} | MEMBER |
ClickHouse | ClickHouse | 593,839,376 | 7,133 | alesapin | Fix in PR #9463. Will be available in all fresh patch releases. | 2020-03-03T08:58:32 | 2020-03-03T08:58:32 | {
"heart": 1
} | MEMBER |
ClickHouse | ClickHouse | 593,879,452 | 9,483 | KochetovNicolai | @alexey-milovidov please don't merge till perftests are finished. | 2020-03-03T10:32:12 | 2020-03-03T10:32:12 | {} | MEMBER |
ClickHouse | ClickHouse | 592,789,351 | 9,421 | den-crane | >Is it true
True.
There are some conversion utility in CH source code. But it's unsupported for a long time and has some restrictions (converts all tables, does not support replicated tables ....). | 2020-02-29T00:22:36 | 2020-02-29T00:22:36 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 592,972,804 | 9,442 | alexey-milovidov | @amosbird Yes, it is absolutely unclear how does it help vectorization. Akazz invited us to guess :) | 2020-02-29T17:39:47 | 2020-02-29T17:39:47 | {} | MEMBER |
ClickHouse | ClickHouse | 593,318,653 | 9,458 | blinkov | @alexey-milovidov please uninstall `fonts-liberation` and `fonts-liberation2` deb packages, reboot your PC and provide screenshots of same content after that. | 2020-03-02T09:55:38 | 2020-03-02T09:55:38 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 593,462,641 | 2,581 | alesapin | @vasyaabr Проблема исправлена в PR #2863. Если что-то всё ещё воспроизводится, то можно переоткрыть этот issue. | 2020-03-02T15:35:51 | 2020-03-02T15:36:00 | {} | MEMBER |
ClickHouse | ClickHouse | 593,923,918 | 9,323 | alexey-milovidov | Still need to fix two queries in perf test. | 2020-03-03T12:23:09 | 2020-03-03T12:23:09 | {} | MEMBER |
ClickHouse | ClickHouse | 592,545,446 | 9,387 | nikitamikhaylov | Conflicts in 20.2 && 20.1
| 2020-02-28T14:48:17 | 2020-02-28T14:52:42 | {} | MEMBER |
ClickHouse | ClickHouse | 592,674,270 | 9,435 | alesapin | Doesn't affect failed tests | 2020-02-28T18:55:43 | 2020-02-28T18:55:43 | {} | MEMBER |
ClickHouse | ClickHouse | 592,674,471 | 9,427 | alesapin | Tests are OK. | 2020-02-28T18:56:17 | 2020-02-28T18:56:17 | {} | MEMBER |
ClickHouse | ClickHouse | 592,658,368 | 9,425 | den-crane | Create a view without ORDER BY
and `any(QHour) as QHour`
It's impossible to pushdown `any(QHour)`
```
insert into source(TIMESTAMP) select toDateTime('2019-01-01 00:00:00')+number from numbers(10000000);
SELECT *
FROM
(
SELECT
CELL,
QHour,
toStartOfFifteenMinutes(TIMESTA... | 2020-02-28T18:16:10 | 2020-02-28T20:58:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 592,731,410 | 9,451 | CurtizJ | > Do I understand correctly that the order was non-deterministic because the order of iteration of std::unordered_map depends on the insertion order when elements are hashed to the same cells?
Yes. | 2020-02-28T21:14:25 | 2020-02-28T21:14:25 | {} | MEMBER |
ClickHouse | ClickHouse | 593,274,390 | 9,412 | alesapin | Tests are OK. | 2020-03-02T08:12:38 | 2020-03-02T08:12:38 | {} | MEMBER |
ClickHouse | ClickHouse | 593,374,883 | 9,454 | alexey-milovidov | > collect2: fatal error: ld terminated with signal 9 [Killed]
OOM in build. | 2020-03-02T12:18:19 | 2020-03-02T12:18:19 | {} | MEMBER |
ClickHouse | ClickHouse | 593,713,610 | 9,472 | alexey-milovidov | This partially reverts #8765 | 2020-03-03T01:08:55 | 2020-03-03T01:08:55 | {} | MEMBER |
ClickHouse | ClickHouse | 593,758,547 | 9,478 | den-crane | You can use Replacing or Collapsing or Aggregating mergeTree
https://gist.github.com/den-crane/d03524eadbbce0bafa528101afa8f794 | 2020-03-03T04:13:15 | 2020-03-03T04:13:31 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 593,824,143 | 6,898 | ilyakarol | cat /etc/clickhouse-server/config.d/ssl.xml:
```xml
<yandex>
<https_port>8443</https_port>
<tcp_port_secure>9440</tcp_port_secure>
<openSSL>
<server>
<certificateFile>/etc/ssl/clickhouse/clickhouse-production.crt</certificateFile>
<privateKeyFile>/etc/ssl/clickhouse/c... | 2020-03-03T08:18:24 | 2020-03-03T08:18:24 | {} | NONE |
ClickHouse | ClickHouse | 593,907,427 | 9,481 | 4ertus2 | It's also possible to remove this code: https://github.com/ClickHouse/ClickHouse/blob/master/dbms/src/Interpreters/SyntaxAnalyzer.cpp#L901
| 2020-03-03T11:37:11 | 2020-03-03T11:37:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 593,912,707 | 9,378 | filimonov | The problem reproduces using your tests case.
Missing blocks correlate well with those exceptions:
```
2020.03.03 11:29:15.628156 [ 20 ] {} <Error> void DB::StorageKafka::threadFunc(): Code: 225, e.displayText() = DB::Exception: ZooKeeper session has been expired.: while write prefix to view default.mv, Stack tra... | 2020-03-03T11:52:06 | 2020-03-03T12:15:07 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 592,541,272 | 9,432 | akuzm | ```
0.4875 0.4637 -0.049 [0,0.014,0.028,0.034] SELECT count() FROM numbers(100000000) WHERE NOT ignore(exp10(toFloat32(number)))
0.0262 0.0235 -0.104 [0.013,0.077,0.122,0.141] SELECT count() FROM numbers_mt(100000000) WHERE NOT ignore(exp10(toFloat32(number)))
0.4152 0.4175 0.005 [0.001,0.007,0.018,0.021] ... | 2020-02-28T14:38:25 | 2020-02-28T14:38:25 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 592,549,504 | 9,322 | nikitamikhaylov | Conflicts in 20.1
| 2020-02-28T14:57:26 | 2020-02-28T14:57:26 | {} | MEMBER |
ClickHouse | ClickHouse | 592,759,264 | 9,261 | CurtizJ | [Result](https://clickhouse-test-reports.s3.yandex.net/9261/f3a22043f34e2baa1b2a4acdf741bafb69ead78b/performance_comparison/report.html) of performance comparison test.
Old, s | New, s | Relative difference (new - old)/old | Randomization distribution quantiles [5%, 50%, 95%, 99%] | Test | Query
-- | -- | -- | -- |... | 2020-02-28T22:31:06 | 2020-02-28T22:31:06 | {
"hooray": 1
} | MEMBER |
ClickHouse | ClickHouse | 592,973,997 | 9,450 | alexey-milovidov | `ReadTimeout: HTTPSConnectionPool(host='api.github.com', port=443): Read timed out. (read timeout=15)`
Let's assume it's Ok.
GitHub was very unstable last week. | 2020-02-29T17:53:44 | 2020-02-29T17:54:01 | {} | MEMBER |
ClickHouse | ClickHouse | 593,511,428 | 2,345 | CurtizJ | Analysis of index is not significant in these queries.
This case was optimized in several steps. Firstly there was implemented optimization of sorting with using order key, but difference between queries still existed. Finally performance of nested subqueries was fixed in #9235. This PR eliminates unnecessery reading ... | 2020-03-02T17:12:16 | 2020-03-02T17:12:16 | {} | MEMBER |
ClickHouse | ClickHouse | 592,718,709 | 9,451 | alexey-milovidov | Do I understand correctly that the order was non-deterministic because the order of iteration of std::unordered_map depends on the insertion order when elements are hashed to the same cells? | 2020-02-28T20:41:11 | 2020-02-28T20:41:11 | {} | MEMBER |
ClickHouse | ClickHouse | 592,973,935 | 9,450 | alexey-milovidov | query_profiler - failed in master
Build for AArch64 was timed out, need to investigate... | 2020-02-29T17:53:06 | 2020-02-29T17:53:06 | {} | MEMBER |
ClickHouse | ClickHouse | 593,383,030 | 9,454 | alexey-milovidov | It's slightly better than using non-const version and applying `materialize` function - I consider the task is done. | 2020-03-02T12:42:04 | 2020-03-02T12:42:11 | {} | MEMBER |
ClickHouse | ClickHouse | 593,448,936 | 9,300 | alexey-milovidov | Discussed in TG chat. | 2020-03-02T15:08:47 | 2020-03-02T15:08:47 | {} | MEMBER |
ClickHouse | ClickHouse | 593,921,593 | 9,472 | alexey-milovidov | https://nda.ya.ru/t/9dD-CTAI3VucGG | 2020-03-03T12:16:38 | 2020-03-03T12:16:38 | {} | MEMBER |
ClickHouse | ClickHouse | 592,546,387 | 9,432 | akuzm | ```
5.902 [4,134,795] SoftPageFaults SELECT count() FROM numbers_mt(100000000) WHERE NOT ignore(exp10(toFloat64(number)))
3 [1,1,4] string:1426#Poco::ConsoleChannel::log(Poco::Message const&) SELECT count() FROM numbers_mt(100000000) WHERE NOT ignore(exp10(toFloat64(number)))
3 [1,1,4] OwnSplitChannel.cpp:44#DB::Own... | 2020-02-28T14:50:28 | 2020-02-28T14:50:28 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 592,973,484 | 9,442 | alexey-milovidov | BTW, I don't see the relevant changes in performance test. We can merge anyway. | 2020-02-29T17:48:07 | 2020-02-29T17:48:07 | {} | MEMBER |
ClickHouse | ClickHouse | 593,004,940 | 9,449 | vitlibar | > Test required.
done | 2020-02-29T23:20:19 | 2020-02-29T23:20:19 | {} | MEMBER |
ClickHouse | ClickHouse | 593,413,206 | 9,461 | alexey-milovidov | #9465 | 2020-03-02T13:54:25 | 2020-03-02T13:54:25 | {} | MEMBER |
ClickHouse | ClickHouse | 593,717,352 | 9,472 | alexey-milovidov | There were a few bugs:
1. The check for signal overrun is not done before collecting a stack trace, that defeats the purpose of this check:
https://github.com/ClickHouse/ClickHouse/pull/8765/files#diff-1f307cce4cc43b590c8b9c2023ab213fR35
This leads to deadlock.
2. Bad calculation of the next profile step:
... | 2020-03-03T01:22:26 | 2020-03-03T01:54:33 | {} | MEMBER |
ClickHouse | ClickHouse | 592,539,771 | 9,262 | nikitamikhaylov | Conflicts in 20.2
| 2020-02-28T14:35:02 | 2020-02-28T14:35:02 | {} | MEMBER |
ClickHouse | ClickHouse | 592,833,777 | 9,442 | amosbird | It would be ideal to see the changed assembly and with a comment about how this helps vectorization. BTW is it compiler related? | 2020-02-29T03:02:10 | 2020-02-29T03:02:10 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 592,919,704 | 2,719 | deniskv | Any updates? | 2020-02-29T08:28:21 | 2020-02-29T08:28:21 | {} | NONE |
ClickHouse | ClickHouse | 592,972,932 | 9,442 | alexey-milovidov | > Unit tests — fail: 1, passed: 5205
It's my mistake that was in master.
We can merge this PR after adding a comment... | 2020-02-29T17:41:14 | 2020-02-29T17:41:14 | {} | MEMBER |
ClickHouse | ClickHouse | 593,693,549 | 9,459 | alexey-milovidov | > Compatibility check — Old ubuntu:12.04 failed
Don't see test results. Is it a CI glitch? | 2020-03-03T00:06:39 | 2020-03-03T00:06:54 | {} | MEMBER |
ClickHouse | ClickHouse | 593,739,511 | 9,049 | alexey-milovidov | Should continue after #9472 | 2020-03-03T02:49:52 | 2020-03-03T02:49:52 | {} | MEMBER |
ClickHouse | ClickHouse | 593,918,665 | 9,472 | alexey-milovidov | The new test "memory profiler" is only for release build. | 2020-03-03T12:09:01 | 2020-03-03T12:09:01 | {} | MEMBER |
ClickHouse | ClickHouse | 592,598,115 | 9,246 | KochetovNicolai | It should be partly fixed after #8929 | 2020-02-28T16:47:11 | 2020-02-28T16:47:11 | {} | MEMBER |
ClickHouse | ClickHouse | 592,681,942 | 9,393 | alexey-milovidov | This is wrong. | 2020-02-28T19:14:10 | 2020-02-28T19:14:10 | {} | MEMBER |
ClickHouse | ClickHouse | 592,794,020 | 9,428 | den-crane | DateTime -- Int32 -- number of seconds from 1970-01-01 00:00:00 UTC.
2^ 32 / 3600 / 24 /365 ~ 136 years
1970 + 136 = 2106
So !!! Int32 able to store only a period of 136 years -- 1970 -- 2106
```
(2201-1970)*3600*24*365 = 7284816000 seconds bigger than 2^32 (4294967296)
```
```
SELECT toUInt32(7284816... | 2020-02-29T00:42:11 | 2020-02-29T01:07:21 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 593,703,025 | 9,358 | alexey-milovidov | Perfect, thank you! | 2020-03-03T00:27:42 | 2020-03-03T00:27:42 | {} | MEMBER |
ClickHouse | ClickHouse | 592,542,508 | 9,432 | akuzm | https://clickhouse-test-reports.s3.yandex.net/0/ae8a70138730ed4f4bfea01982305ccb2f28b4b2/performance_comparison/report.html#5
The worst tests are codecs and hashes, as before. I think I'll leave them as is for now, and first try to fix the very slow tests and the tests that fail to run, so that we can switch to comp... | 2020-02-28T14:41:13 | 2020-02-28T14:41:13 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 592,549,022 | 9,345 | nikitamikhaylov | Conflicts in 20.1
| 2020-02-28T14:56:24 | 2020-02-28T14:56:24 | {} | MEMBER |
ClickHouse | ClickHouse | 592,940,490 | 9,393 | ghost | @alexey-milovidov if it's not urgent, i'm willing to make another pull request. it may take me some time. | 2020-02-29T12:29:45 | 2020-02-29T12:29:45 | {} | NONE |
ClickHouse | ClickHouse | 592,976,709 | 9,420 | alexey-milovidov | It didn't finish. But it's definitely caused by the issues of GitHub API. | 2020-02-29T18:21:40 | 2020-02-29T18:21:40 | {} | MEMBER |
ClickHouse | ClickHouse | 593,454,693 | 9,466 | alesapin | Merge test fix without tests. | 2020-03-02T15:20:01 | 2020-03-02T15:20:01 | {} | MEMBER |
ClickHouse | ClickHouse | 593,724,877 | 9,472 | alexey-milovidov | Ok, now it works perfectly:
```
SELECT ignore(groupArray(number)) FROM numbers(10000000)
```
```
SELECT
arrayStringConcat(arrayMap(x -> demangle(addressToSymbol(x)), trace), '\n') AS trace,
formatReadableSize(sum(size) AS s)
FROM system.trace_log
WHERE (event_date = today()) AND (query_id = 'd464df... | 2020-03-03T01:53:19 | 2020-03-03T01:53:19 | {} | MEMBER |
ClickHouse | ClickHouse | 593,728,336 | 9,472 | alexey-milovidov | > Remove weak aliases from musl
It didn't help.
Now I think the issue is the lack of unwind tables (CFI) in asm sources in Musl. | 2020-03-03T02:07:16 | 2020-03-03T02:07:16 | {} | MEMBER |
ClickHouse | ClickHouse | 593,883,598 | 9,389 | filimonov | Let's try newer binary build instead. | 2020-03-03T10:41:55 | 2020-03-03T10:41:55 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 593,889,272 | 9,484 | filimonov | ```
docker run -it --rm yandex/clickhouse-stateless-test ldd /usr/local/lib/libclickhouseodbc.so
linux-vdso.so.1 (0x00007fff35f8e000)
libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f1583a1a000)
libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007f1583816000)
librt.so.1 => /lib/x86_64-linux... | 2020-03-03T10:54:28 | 2020-03-03T10:54:28 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 593,930,442 | 9,485 | filimonov | You can download & unpack clickhouse-odbc on ubuntu like that
```
wget "https://github.com/ClickHouse/clickhouse-odbc/releases/download/v1.1.4.20200302/clickhouse-odbc-1.1.4-Linux.tar.gz"
tar xvzf clickhouse-odbc-1.1.4-Linux.tar.gz
```
After that you may need to install it like that:
```
apt-get install unix... | 2020-03-03T12:39:33 | 2020-03-03T12:41:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 592,653,563 | 9,426 | den-crane | Need steps to reproduce. Table DDL, sample data. SQL in text form , not as picture | 2020-02-28T18:03:53 | 2020-02-28T18:03:53 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 592,723,755 | 8,782 | alexey-milovidov | @KochetovNicolai There is no visible difference. | 2020-02-28T20:54:56 | 2020-02-28T20:54:56 | {} | MEMBER |
ClickHouse | ClickHouse | 592,725,669 | 9,433 | alexey-milovidov | Performance test Ok. | 2020-02-28T20:59:51 | 2020-02-28T20:59:51 | {} | MEMBER |
ClickHouse | ClickHouse | 593,109,813 | 9,406 | byx313 | > No.
>
> Currently `INSERT INTO SELECT` works in one thread (insert part).
> In the next stable release (20.2) will be implemented **Parallel INSERT in INSERT SELECT query** #8166 and a setting `max_insert_threads`.
>
> But still where is no ability and plans for ability to control insert rate in MB/s or rows/s... | 2020-03-01T15:29:16 | 2020-03-01T15:29:16 | {} | NONE |
ClickHouse | ClickHouse | 593,488,144 | 9,442 | Akazz | First off, here's the request that I am testing against:
```
SELECT count() FROM (SELECT materialize(toUInt16(1)) AS x1, materialize(toUInt32(1)) AS x2 FROM numbers(20000000000)) WHERE NOT ignore(and(x1,x2))
```
I found no significant difference in performance depending on compiler (GCC 9.2.1 vs Clang 9.0.0), altho... | 2020-03-02T16:25:15 | 2020-03-02T17:23:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 593,725,717 | 9,472 | alexey-milovidov | We have our own clock_nanosleep from Musl. | 2020-03-03T01:56:40 | 2020-03-03T01:56:40 | {} | MEMBER |
ClickHouse | ClickHouse | 593,734,707 | 9,472 | alexey-milovidov | And this solves the issue: 8d679b5 | 2020-03-03T02:30:37 | 2020-03-03T02:30:37 | {} | MEMBER |
ClickHouse | ClickHouse | 592,724,821 | 8,594 | alexey-milovidov | (rebased but boost::shared_ptr remains) | 2020-02-28T20:57:38 | 2020-02-28T20:57:38 | {} | MEMBER |
ClickHouse | ClickHouse | 592,767,499 | 9,393 | alexey-milovidov | Specializations of `insert` method in PODArray with type U that is different than T (and also probably, not trivially copyable) require a loop with element by element assignment instead of memcpy.
kudos @Akazz @KochetovNicolai | 2020-02-28T22:57:56 | 2020-02-28T22:59:20 | {} | MEMBER |
ClickHouse | ClickHouse | 592,792,124 | 9,428 | ggchangan | @den-crane Thanks very much for your response. However I still can't get your point. Can you give more details? When the Int32 overflow happens? How the Int32 overflow Happens? Can you give some source code to explain this? | 2020-02-29T00:34:35 | 2020-02-29T00:34:35 | {} | NONE |
ClickHouse | ClickHouse | 592,976,396 | 9,446 | alexey-milovidov | Strange performance comparison results.
I cannot say that anything is wrong, let's merge. | 2020-02-29T18:18:45 | 2020-02-29T18:18:45 | {} | MEMBER |
ClickHouse | ClickHouse | 593,402,449 | 7,133 | alesapin | Trying to fix here https://github.com/ClickHouse/ClickHouse/pull/9463/files. | 2020-03-02T13:30:24 | 2020-03-02T13:30:24 | {} | MEMBER |
ClickHouse | ClickHouse | 592,546,944 | 9,011 | nikitamikhaylov | Conflicts in 20.1
| 2020-02-28T14:51:46 | 2020-02-28T14:51:46 | {} | MEMBER |
ClickHouse | ClickHouse | 592,976,084 | 9,447 | alexey-milovidov | Integration tests are to be updated? | 2020-02-29T18:15:27 | 2020-02-29T18:15:27 | {} | MEMBER |
ClickHouse | ClickHouse | 593,500,517 | 9,442 | Akazz | I've looked into it a little bit deeper and isolated the problem to be with ... `<` - the comparison operator! Clearly, our existing implementation of `PODArray::iterator::operator<()` prevents compiler from employing further optimizations. See the new patch!
GCC 9.2.1:
```
│ *it = data[i];
... | 2020-03-02T16:49:58 | 2020-03-02T17:42:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 593,726,877 | 9,472 | alexey-milovidov | Almost obviously, the issue with `lld` linker has the same cause: https://github.com/ClickHouse/ClickHouse/pull/9049 | 2020-03-03T02:01:21 | 2020-03-03T02:01:21 | {} | MEMBER |
ClickHouse | ClickHouse | 593,822,420 | 6,898 | ilyakarol | Errors appear again, Clickhouse version `19.16.13.54`.
It is very weird, because we have `<disableProtocols>sslv2,sslv3,tlsv1,tlsv1_1</disableProtocols>` and is says `tlsv1 alert decrypt error` | 2020-03-03T08:13:28 | 2020-03-03T08:13:28 | {} | NONE |
ClickHouse | ClickHouse | 592,534,758 | 9,351 | nikitamikhaylov | Conflicts in 20.2 && 20.1
| 2020-02-28T14:23:19 | 2020-02-28T14:55:57 | {} | MEMBER |
ClickHouse | ClickHouse | 592,767,811 | 9,393 | alexey-milovidov | This can be improved... We just need more specializations.
Do you want to make another pull request on top of my changes? | 2020-02-28T22:59:04 | 2020-02-28T22:59:04 | {} | MEMBER |
ClickHouse | ClickHouse | 593,094,665 | 6,892 | alexey-milovidov | It looks almost ready, let's finish. | 2020-03-01T13:02:40 | 2020-03-01T13:02:40 | {} | MEMBER |
ClickHouse | ClickHouse | 593,345,773 | 9,357 | martenlindblad | The build pass this step now. | 2020-03-02T10:59:08 | 2020-03-02T10:59:08 | {} | NONE |
ClickHouse | ClickHouse | 593,428,941 | 8,989 | ssserj | @alexey-milovidov please, try to fix issue #8890 in 20.2 release. Can't work properly (expire policy doesn't work) with external dictionaries with `source_type` DBMS::MySQL at now. | 2020-03-02T14:28:29 | 2020-03-02T14:28:29 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 592,972,405 | 9,393 | alexey-milovidov | Not urgent at all. Yes, it will be appreciated! | 2020-02-29T17:35:42 | 2020-02-29T17:35:42 | {} | MEMBER |
ClickHouse | ClickHouse | 593,106,242 | 9,456 | den-crane | No.
You can change `order by` -- it will not change primary key index, but will change merge collapsing/aggregating behavior. | 2020-03-01T14:56:22 | 2020-03-01T14:56:58 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 593,108,474 | 9,423 | den-crane | 1) You should always use sum / group by when querying SummingMergeTree.
Please provide an example (DDL/Dataset/Query) of bad performance.
2) You can use AggregatingFunctions with SummingMergeTree. | 2020-03-01T15:16:39 | 2020-03-01T15:16:39 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 593,663,267 | 9,442 | danlark1 | The problem is in boost::iterator https://gcc.godbolt.org/z/qgwFCF
```
<source>:44:9: remark: loop not vectorized: could not determine number of loop iterations [-Rpass-analysis=loop-vectorize]
for (; it < span_end; ++it)
^
``` | 2020-03-02T22:47:38 | 2020-03-02T22:48:57 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 593,823,206 | 9,421 | alesapin | If you have any related problems, just reopen this issue. | 2020-03-03T08:15:48 | 2020-03-03T08:15:48 | {} | MEMBER |
ClickHouse | ClickHouse | 592,657,173 | 8,303 | nikitamikhaylov | There is no each_update_finish_timeout_seconds setting in config. We got rid about it after discussion. | 2020-02-28T18:13:04 | 2020-02-28T18:13:04 | {} | MEMBER |
ClickHouse | ClickHouse | 592,761,742 | 9,393 | ghost | what's wrong? | 2020-02-28T22:38:31 | 2020-02-28T22:38:31 | {} | NONE |
ClickHouse | ClickHouse | 592,913,371 | 5,295 | ilyakarol | The same problem, 19.16.13.54 | 2020-02-29T07:19:51 | 2020-02-29T07:19:51 | {} | NONE |
ClickHouse | ClickHouse | 592,973,786 | 9,451 | alexey-milovidov | query_profiler - also failed in master
00965_live_view_watch_heartbeat - flap for unknown reason
Integration tests (asan) - CI cannot run the cluster with docker. | 2020-02-29T17:51:38 | 2020-02-29T17:51:38 | {} | MEMBER |
ClickHouse | ClickHouse | 593,109,515 | 9,406 | den-crane | No.
Currently `INSERT INTO SELECT` works in one thread (insert part).
In the next stable release (20.2) will be implemented **Parallel INSERT in INSERT SELECT query** #8166 and a setting `max_insert_threads`.
But still where is no ability and plans for ability to control insert rate in MB/s or rows/s.
As w... | 2020-03-01T15:26:17 | 2020-03-01T15:26:39 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 593,898,498 | 9,473 | filimonov | Related (maybe bit more complicated) ability to stop printing the big output to console when Ctrl+C is pressed.
Use case:
```
SELECT * FROM huge_table_on_remote_server;
<console get flooded by table data>
^C <- kill get send to the server
<console get flooded by table data for a very long time, untill server... | 2020-03-03T11:12:27 | 2020-03-03T11:12:27 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 592,543,041 | 9,428 | den-crane | Int32 overflow | 2020-02-28T14:42:31 | 2020-02-28T14:42:31 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 591,773,686 | 9,361 | jangorecki | > A table will be without any order key. (Order by 1).
By not having any order, you mean that it retains the order of the original data during insert, or data are inserted in random order? @den-crane
> > > ORDER BY (id0) -- the worst idea as it can be -- NEVER EVER DO THAT.
> > > why?
>
> Because the idea be... | 2020-02-27T04:25:54 | 2020-03-01T10:39:29 | {} | NONE |
ClickHouse | ClickHouse | 593,382,133 | 9,454 | alexey-milovidov | Up to 13 times faster.
Before:
```
│ Disassembly of section .text: ... | 2020-03-02T12:39:22 | 2020-03-02T12:39:22 | {} | MEMBER |
ClickHouse | ClickHouse | 593,382,492 | 9,454 | alexey-milovidov | Actually 30 times faster (on synthetic test).
It's vectorized (though I'm not sure that it's done in the best way). | 2020-03-02T12:40:21 | 2020-03-02T12:40:21 | {} | MEMBER |
ClickHouse | ClickHouse | 593,400,815 | 9,458 | alexey-milovidov | > please uninstall fonts-liberation and fonts-liberation2 deb packages
This is not an option. I have tested on two machines and font rendering is bad on both of them. If the "Liberation" font has low quality, we just should not use it instead of insisting users to uninstall it. | 2020-03-02T13:26:55 | 2020-03-02T13:27:15 | {} | MEMBER |
ClickHouse | ClickHouse | 593,719,032 | 9,472 | alexey-milovidov | QueryProfiler test was broken in https://github.com/ClickHouse/ClickHouse/pull/9433
For unknown reason, libunwind cannot unwind the stack above the signal handler if signal has come to `clock_nanosleep` function. Probably something is wrong with unwind tables or linker.
```
┌─trace─────────────────────────────────... | 2020-03-03T01:29:18 | 2020-03-03T01:29:18 | {} | MEMBER |
ClickHouse | ClickHouse | 593,719,819 | 9,472 | alexey-milovidov | ```
$ grep memory_profiler *.sql
$ grep memory_profiler *.sh
```
Also there was no test for memory profiler. | 2020-03-03T01:32:25 | 2020-03-03T01:47:38 | {} | MEMBER |
ClickHouse | ClickHouse | 593,927,988 | 9,484 | qoega | Actually it can't be tested in CI. | 2020-03-03T12:34:07 | 2020-03-03T12:34:07 | {} | MEMBER |
ClickHouse | ClickHouse | 607,363,820 | 2,245 | blinkov | @StarWix, do you have any further questions? | 2020-04-01T16:46:48 | 2020-04-01T16:46:48 | {} | CONTRIBUTOR |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.