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 | 539,190,728 | 7,213 | azat | >Actually looking into this more and more, and I cannot verify that this will change in-memory/on-disk representation, am I missing something?
Hm, changing for other types `HashValueType` to `uint64` does not break reading HLL from the disk, will you mind if I will change this?
Tested with:
```
create table s... | 2019-10-07T20:30:03 | 2019-10-07T20:30:03 | {} | MEMBER |
ClickHouse | ClickHouse | 539,205,094 | 7,196 | filimonov | > It is logical how it CH works, but, I suppose, it would be more useful if query_duration_ms (or another one) would measure real time of query execution which period is from step 5 to 6 in your list.
That would not be true.
Query was executed with a speed limited by client sending data, but it was executed all ... | 2019-10-07T21:08:36 | 2019-10-07T21:12:57 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,249,967 | 7,213 | alexey-milovidov | uniqCombined has three states:
1. Small flat array.
2. Hash table.
3. HLL.
The sizes of these data structures are aligned to each other. Hash table (2) can grow only up to the size of HLL (3) and is converted to HLL when becomes too big.
Flat array and hash table stores hash values directly. But HLL does not s... | 2019-10-07T23:38:55 | 2019-10-08T00:11:07 | {} | MEMBER |
ClickHouse | ClickHouse | 539,472,718 | 6,874 | alexey-milovidov | @Artemiyfly wants to do this task. | 2019-10-08T11:36:32 | 2019-10-08T11:36:32 | {} | MEMBER |
ClickHouse | ClickHouse | 539,483,672 | 7,224 | tavplubix | JFYI: Another alternative is using `system.query_log` table to get various information about failed (`type=3 or type=4`) queries without parsing server logs for exceptions. | 2019-10-08T12:12:41 | 2019-10-08T12:12:41 | {} | MEMBER |
ClickHouse | ClickHouse | 539,491,254 | 7,144 | den-crane | I don't see a problem with Check. Maybe it's a false positive. Please merge it. I'll fix it later if real issue after rendering into HTML. | 2019-10-08T12:34:52 | 2019-10-08T12:34:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 538,918,051 | 6,661 | filimonov | > @abyss7 during my tests Kafka topic didn't contain the same data as I inserted manually. It's not a big problem for me, I just used temporary table (insert in a temporary table and move data manually) in my scenario, just wanted to avoid one intermediate step.
Can you describe the exact steps with expected and rea... | 2019-10-07T09:32:33 | 2019-10-07T09:32:33 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 538,918,530 | 7,194 | alesapin | Manually cherry-picked to 19.15. | 2019-10-07T09:33:32 | 2019-10-07T09:33:32 | {} | MEMBER |
ClickHouse | ClickHouse | 539,191,169 | 6,448 | kreuzerkrieg | > I have read almost all changes and this PR is generally Ok.
Except the fact the GCC does not compile, for now. I'm waiting for the build to fail to see the preprocessor output | 2019-10-07T20:31:08 | 2019-10-07T20:31:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,461,728 | 6,897 | tavplubix | When inserting we don't know the total amount of rows until all of them are parsed, so errors ratio is calculated for rows we've seen. Probably, we could parse all rows just to calculate errors ratio for the whole batch, but it doesn't seem to be ok for large inserts. Because of it, `input_format_allow_errors_ratio` sh... | 2019-10-08T10:59:36 | 2019-10-08T10:59:36 | {} | MEMBER |
ClickHouse | ClickHouse | 539,605,843 | 7,229 | den-crane | And how are going to set `<shard>` and `<layer>` ?
I just use `chef` and generate macros.xml using internal cmdb which knows what node is which shard, and substitutes <replica with hostname. | 2019-10-08T16:55:12 | 2019-10-08T16:55:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,674,702 | 7,221 | azat | Hm, but don't we need care only about on-disk representation? (on disk it should *not greater then* 65546 bytes -- `(1<<14)*4+8+2`) | 2019-10-08T19:52:20 | 2019-10-08T20:08:49 | {} | MEMBER |
ClickHouse | ClickHouse | 539,164,102 | 7,113 | alexey-milovidov | @akuzm You can add this example as a test. | 2019-10-07T19:20:15 | 2019-10-07T19:20:15 | {} | MEMBER |
ClickHouse | ClickHouse | 539,459,624 | 7,136 | ghost | Version 19.16.1.1 master is also affected.
```
2019.10.08 18:46:16.846316 [ 23 ] {33ff299f-a32f-4ba8-93f6-aae3e0800bb7} <Debug> executeQuery: (from 127.0.0.1:52736) SELECT ccc FROM (SELECT 1 AS ccc UNION ALL SELECT * FROM (SELECT 2 AS ccc) INNER JOIN (SELECT 2 AS ccc) USING (ccc)) WHERE ccc > 1
2019.10.08 18:46:16.8... | 2019-10-08T10:52:36 | 2019-10-08T10:52:36 | {} | NONE |
ClickHouse | ClickHouse | 539,672,742 | 7,221 | azat | >The size of hash table with max 2^14 elements will contain 2^15 cells of 32bit size:
>```
>>> 4 * 2 * 2 ** 14
>131072
>```
Hm, did not though about that fact that `HashTable<unsigned int>(1<<14)` != `HashTable<unsigned long>(1<<13)`
Also I checked this in debugger, and got this:
- `1<<13` with `String`: `... | 2019-10-08T19:47:16 | 2019-10-08T19:47:16 | {} | MEMBER |
ClickHouse | ClickHouse | 539,676,346 | 6,826 | alexey-milovidov | What is the status of this PR? Why don't we see the hang in unit tests in CI? Is it already fixed and can be merged? | 2019-10-08T19:56:19 | 2019-10-08T19:56:19 | {} | MEMBER |
ClickHouse | ClickHouse | 539,174,458 | 7,213 | alexey-milovidov | Ok. The test: `SELECT uniqCombined(number) FROM numbers_mt(10000000000)`
will not be included in CI, because it's slow. | 2019-10-07T19:48:13 | 2019-10-07T19:48:13 | {} | MEMBER |
ClickHouse | ClickHouse | 539,533,390 | 7,223 | den-crane | eventually 255 partitions will be merged into 75.
>toUInt32(intDiv(start_time, 3600000) * 3600)
has no sense, it's the same as `intDiv(start_time, 3600000)`
run the query with debug (set send_logs_level='debug') against first and 3d tables (format Null -- to hide the data in the result), and provide an outpu... | 2019-10-08T14:13:09 | 2019-10-08T14:13:09 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,628,383 | 7,205 | amosbird | OK, now I see that user `clickhouse` exists for dropping privileges. Perhaps we can apply it to the log directory too. nginx creates the log file before dropping privileges to `http` user so that the log folder is only writable to root. | 2019-10-08T17:52:19 | 2019-10-08T17:52:19 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 538,923,104 | 7,187 | filimonov | > I'm pretty sure that DoubleDelta does in fact compress the data
It is.
> I might switch this over to a (Delta, LZ4) anyways, but it's still a very small column so it doesn't matter much.
For now i would you recommend to do that (you can also consider T64) - as DoubleDelta is extremely slow now. :\
That ... | 2019-10-07T09:45:22 | 2019-10-07T09:49:45 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,236,997 | 7,213 | azat | I've prepared another [patch](https://github.com/ClickHouse/ClickHouse/compare/master...azat-archive:uniqCombined-use_64_bit_hash?expand=1) that will force `uniqCombined` to use 64-bit hash, once this will be resolved | 2019-10-07T22:49:13 | 2019-10-07T22:49:13 | {} | MEMBER |
ClickHouse | ClickHouse | 539,288,453 | 6,897 | slimtom95 | Hi @tavplubix
With delimitter `\n`, now I have only one concern about `input_format_allow_errors_ratio`
> Behavior: the error ratio is not a error ratio in this BATCH but in the records we've SEEN.
If we use `input_format_allow_errors_ratio` alone(and less than 1 of course), any bad records at the front wou... | 2019-10-08T02:44:06 | 2019-10-08T02:44:06 | {} | NONE |
ClickHouse | ClickHouse | 539,397,302 | 7,107 | blinkov | merged in #7220 | 2019-10-08T07:59:35 | 2019-10-08T07:59:35 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,628,106 | 5,438 | clemrodriguez | @alexey-milovidov
Hi Alexey,
I made a forced push on my local branch from my fork (which wasn't the pull request branch). Just after that I saw that the PR is closed.
As I'm not used to Git, I'm upset o... | 2019-10-08T17:51:39 | 2019-10-08T17:51:39 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,641,809 | 7,219 | alexey-milovidov | This is the better link: https://clickhouse.github.io/clickhouse-presentations/meetup30/new_features/#12 | 2019-10-08T18:26:08 | 2019-10-08T18:26:08 | {
"+1": 1
} | MEMBER |
ClickHouse | ClickHouse | 538,917,206 | 6,526 | filimonov | 19.11.8 is also affected.
Instead of restart DETACH/ATTACH can be used to make the column disappear. | 2019-10-07T09:30:27 | 2019-10-07T09:30:27 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,112,929 | 7,096 | akuzm | https://github.com/ClickHouse/ClickHouse/pull/7208 | 2019-10-07T17:12:06 | 2019-10-07T17:12:06 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,158,288 | 7,214 | 4ertus2 | Reproduced
```
USE system;
SELECT dummy
FROM one AS A
INNER JOIN
one
ON A.dummy = one.dummy; -- error
SELECT dummy
FROM one
INNER JOIN
one AS A
ON A.dummy = one.dummy; -- OK
``` | 2019-10-07T19:04:08 | 2019-10-07T19:04:08 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,397,170 | 7,105 | blinkov | merged in #7220 | 2019-10-08T07:59:18 | 2019-10-08T07:59:18 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,495,624 | 7,144 | den-crane | OK. I'll try to change `access_rights` link.
INFO:root:Tag <a href="#../access_rights/">Access rights</a>
INFO:root:Link to nowhere: #../access_rights/ | 2019-10-08T12:45:52 | 2019-10-08T12:45:52 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,496,316 | 7,144 | blinkov | I'd try something like ../../operations/access_rights.md | 2019-10-08T12:47:39 | 2019-10-08T12:47:39 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,610,035 | 7,136 | zhang2014 | Ok, I try to reproduce it in my environment and fix it : ) | 2019-10-08T17:05:27 | 2019-10-08T17:05:27 | {
"heart": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,645,671 | 7,216 | alexey-milovidov | Multiple performance tests have timed out but I don't see that it is related to this PR.
```
--input-files /usr/share/clickhouse-test/performance/number_formatting_formats.xml --query-indexes 0 1 2 3 4 5 6 7 8 9 11 12 13 --input-files /usr/share/clickhouse-test/performance/string_sort.xml --query-indexes 0 1 2 6 7 ... | 2019-10-08T18:35:43 | 2019-10-08T18:36:52 | {} | MEMBER |
ClickHouse | ClickHouse | 539,141,242 | 7,187 | den-crane | Yes, the same issue. Sort of feature request for performance improvement. | 2019-10-07T18:21:48 | 2019-10-07T18:21:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,188,498 | 7,213 | alexey-milovidov | > I used toString here to force uint64 type (this looks odd to me that uint64 only for string but this is another story)
Can you point me to specific line in code? | 2019-10-07T20:24:09 | 2019-10-07T20:24:09 | {} | MEMBER |
ClickHouse | ClickHouse | 539,183,836 | 7,213 | azat | >Ok. The test: SELECT uniqCombined(number) FROM numbers_mt(10000000000)
will not be included in CI, because it's slow.
Indeed, but if there were function like `writeAggregation`/`readAggregation` (the same as `finalizeAggregation`) this can be possible
> what is the result of this test on your machine?
I test... | 2019-10-07T20:12:29 | 2019-10-07T20:24:31 | {} | MEMBER |
ClickHouse | ClickHouse | 539,189,127 | 7,213 | azat | >Can you point me to specific line in code?
- all except string
https://github.com/ClickHouse/ClickHouse/blob/15deedb42017f1214da7fae499dec6eeff14dd1a/dbms/src/AggregateFunctions/AggregateFunctionUniqCombined.h#L102
- string:
https://github.com/ClickHouse/ClickHouse/blob/15deedb42017f1214da7fae499dec6eeff14dd1a... | 2019-10-07T20:25:48 | 2019-10-07T20:26:51 | {} | MEMBER |
ClickHouse | ClickHouse | 539,196,590 | 6,448 | alexey-milovidov | Good idea.
PS. The best way is to add new warnings one by one. | 2019-10-07T20:45:24 | 2019-10-07T20:45:24 | {} | MEMBER |
ClickHouse | ClickHouse | 539,637,711 | 7,205 | alexey-milovidov | > Perhaps we can apply it to the log directory too. nginx creates the log file before dropping privileges to http user so that the log folder is only writable to root.
- but we recreate log files when needed;
- and we start as clickhouse user, instead of starting as root and dropping priveleges. | 2019-10-08T18:15:59 | 2019-10-08T18:15:59 | {} | MEMBER |
ClickHouse | ClickHouse | 539,639,442 | 5,438 | alexey-milovidov | Most likely, you have accidentially force-pushed upstream master code to your PR and now it has no difference to upstream (0 files changed). GitHub has automatically closed this PR because it is empty. (Don't worry, nothing wrong happens to our repo).
Do you have your code in your branch? It's better to create anoth... | 2019-10-08T18:20:23 | 2019-10-08T18:20:23 | {} | MEMBER |
ClickHouse | ClickHouse | 539,209,149 | 7,217 | filimonov | You overwrite default profile which also is used by several system tasks (including writing to query_log).
Please try the following:
1) create one more profile, like `system`, which will be not readonly
2) and put the name of that profile in config, here:
https://github.com/ClickHouse/ClickHouse/blob/15deedb42017... | 2019-10-07T21:19:42 | 2019-10-07T21:21:34 | {
"+1": 1
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,231,562 | 7,187 | filimonov | It look like quite easy to workaround with @den-crane proposed option, or (even better speed) with argMin, i.e. something like
```
SELECT id, a, b FROM Xp
WHERE id IN ( SELECT argMin(id, a) FROM Xp )
ORDER BY a ASC
LIMIT 1
```
But will such optimisation be good "in general"?
For example - imagine that id ext... | 2019-10-07T22:30:35 | 2019-10-07T22:31:06 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,265,261 | 7,171 | alexey-milovidov | Only a few changes remain. | 2019-10-08T00:49:16 | 2019-10-08T00:49:16 | {} | MEMBER |
ClickHouse | ClickHouse | 539,323,506 | 7,188 | thyn | doesn't help
count(1) takes 0.9 sec
uniqExact(id) takes 35 sec
count(1) final takes 53 sec
| 2019-10-08T04:55:38 | 2019-10-08T04:55:38 | {} | NONE |
ClickHouse | ClickHouse | 539,391,835 | 7,192 | qiaochaojie | @den-crane It seems like a bug. When I use the command 'alter table xx drop partition(20191001)' , It works. After that, when I use 'alter table xx on cluster xx drop partition(p_20191001)' on other tables. it all works. I do not know why | 2019-10-08T07:46:26 | 2019-10-08T07:46:26 | {} | NONE |
ClickHouse | ClickHouse | 539,646,034 | 7,216 | alexey-milovidov | Performance test is Ok in https://github.com/ClickHouse/ClickHouse/pull/7007, so I merge this PR. | 2019-10-08T18:36:40 | 2019-10-08T18:36:40 | {} | MEMBER |
ClickHouse | ClickHouse | 539,684,804 | 7,221 | alexey-milovidov | If it is resized and then after adding one next element, converted to HLL, it is waste of memory and CPU and should be fixed. | 2019-10-08T20:17:18 | 2019-10-08T20:17:24 | {} | MEMBER |
ClickHouse | ClickHouse | 539,174,623 | 7,213 | alexey-milovidov | @azat what is the result of this test on your machine? | 2019-10-07T19:48:39 | 2019-10-07T19:48:39 | {} | MEMBER |
ClickHouse | ClickHouse | 539,189,382 | 6,448 | alexey-milovidov | I have read almost all changes and this PR is generally Ok. | 2019-10-07T20:26:30 | 2019-10-07T20:26:30 | {} | MEMBER |
ClickHouse | ClickHouse | 539,191,672 | 7,213 | azat | >does not break reading HLL from the disk
But this will break uniqueness property of already added values. Maybe a separate function? Thoughts? | 2019-10-07T20:32:23 | 2019-10-07T20:32:23 | {} | MEMBER |
ClickHouse | ClickHouse | 539,192,292 | 6,448 | kreuzerkrieg | > BTW, I found these changes very useful. Especially: forcing functions to have internal linkage, forcing override...
`override` stuff is far from being completed, it has tons of warnings emitted, I temporary disabled the GCC switch, otherwise it was unbearable. Maybe it is a good idea in the end to merge this speci... | 2019-10-07T20:33:55 | 2019-10-07T20:33:55 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,280,237 | 7,192 | qiaochaojie | @den-crane Thanks for your answer. I drop partition with this 'alter table xx on cluster '{cluster}' drop partition(20191001)'. The system.parts shows that partition field is just like '20191001'. | 2019-10-08T02:04:29 | 2019-10-08T02:04:29 | {} | NONE |
ClickHouse | ClickHouse | 539,519,672 | 7,205 | proller | Why you want to change owner of etc/* directories?
Usually only root can change config files.
Making write access to "clickhouse" user is insecure.
Preprocessed configs was moved from etc to data dir witj this reason. | 2019-10-08T13:42:41 | 2019-10-08T13:42:41 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,528,735 | 7,219 | ifoo | There is documentation on the russian version of the MergeTree page here: https://clickhouse.yandex/docs/ru/operations/table_engines/mergetree/#table_engine-mergetree-multiple-volumes_configure . Unfortunately, it's not in the english version yet. There are also some examples in a recent presentation here: https://html... | 2019-10-08T14:02:40 | 2019-10-08T14:02:40 | {
"+1": 1
} | NONE |
ClickHouse | ClickHouse | 539,650,690 | 6,161 | akuzm | New failure with Null fields: Null is converted to Int64 with get<>, giving an uninitialized result.
This is different from the earlier failure where UInt8 was converted to UInt64 (6960b893fdd0a3147ead7e0125e213988e0916b9)
The test results:
https://clickhouse-test-reports.s3.yandex.net/7231/2e81ad7f335a26ad0949e8a... | 2019-10-08T18:48:57 | 2019-10-08T18:48:57 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,667,061 | 7,221 | alexey-milovidov | It doesn't look correct:
```
For K = 17:
The size of HLL in bytes:
>>> 2 ** 17 * 6 / 8
98304
The size of hash table with max 2^14 elements will contain 2^15 cells of 32bit size:
>>> 4 * 2 * 2 ** 14
131072
And this is greater than the size of HLL.
``` | 2019-10-08T19:31:51 | 2019-10-08T19:32:09 | {} | MEMBER |
ClickHouse | ClickHouse | 539,673,329 | 7,221 | azat | >SELECT intDiv(number, 16384) AS k, uniqCombined(number % 16384) FROM numbers(16384000) GROUP BY k;
And indeed, this one fails after the patch | 2019-10-08T19:48:42 | 2019-10-08T19:48:42 | {} | MEMBER |
ClickHouse | ClickHouse | 539,674,449 | 7,109 | alexey-milovidov | Style check does not show any errors after merge. | 2019-10-08T19:51:39 | 2019-10-08T19:51:39 | {} | MEMBER |
ClickHouse | ClickHouse | 539,685,983 | 7,221 | azat | Okay, thanks, closing | 2019-10-08T20:20:08 | 2019-10-08T20:20:08 | {} | MEMBER |
ClickHouse | ClickHouse | 539,397,273 | 7,106 | blinkov | merged in #7220 | 2019-10-08T07:59:29 | 2019-10-08T07:59:29 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,561,446 | 7,205 | amosbird | > Usually only root can change config files.
I don't understand. Isn't `clickhouse` user supposed to be the dbadmin user? How come it's insecure when user `clickhouse` can rw data?
| 2019-10-08T15:13:32 | 2019-10-08T15:13:32 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 539,640,441 | 7,205 | amosbird | Yeah, it's easier to just have log dir belong to `clickhouse` | 2019-10-08T18:22:55 | 2019-10-08T18:22:55 | {} | COLLABORATOR |
ClickHouse | ClickHouse | 538,990,213 | 7,176 | akuzm | Thanks for the example, I can reproduce the error.
It's an issue with wrong Field type being deduced for some Tuple values, which we also noticed recently. I'm fixing it in this PR: https://github.com/ClickHouse/ClickHouse/pull/7179 | 2019-10-07T12:39:10 | 2019-10-07T12:39:10 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,047,920 | 7,187 | den-crane | @filimonov the issue is not related to CODECs.
```
create table Xp(id Int64, a UInt8, b String) Engine = MergeTree order by id
settings index_granularity=1;
insert into Xp select number,
number%103,
toString(arrayMap(x->sipHash64(number,x), range(50)))
from numbers(10000000);
set max_threa... | 2019-10-07T14:46:48 | 2019-10-07T14:46:48 | {
"+1": 4
} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,110,832 | 7,202 | luck143 | On further experiment i found the following timezones fails to provide expected result
```
Africa/El_Aaiun
Asia/Pyongyang
Pacific/Kwajalein
Pacific/Apia
Pacific/Enderbury
Pacific/Fakaofo
Pacific/Kiritimati
```
on the queries like these :
`select toStartOfDay(createdtimestr , 'Africa/El_Aaiun' ) AS t... | 2019-10-07T17:06:36 | 2019-10-07T17:08:06 | {} | NONE |
ClickHouse | ClickHouse | 539,131,032 | 7,187 | filimonov | Ok, reopened. but is it still the same isuue as in original issue description? | 2019-10-07T17:56:57 | 2019-10-07T17:56:57 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,181,267 | 6,448 | alexey-milovidov | Except "missing-declarations". Probably it'd better be rolled back. | 2019-10-07T20:06:01 | 2019-10-07T20:06:01 | {} | MEMBER |
ClickHouse | ClickHouse | 539,606,756 | 7,205 | alexey-milovidov | Suppose we have some vulnerability in ClickHouse that allow us to write arbitrary files in filesystem. If the configs are writeable by clickhouse user, this will allow to change config using this vulnerability. | 2019-10-08T16:57:25 | 2019-10-08T16:57:25 | {} | MEMBER |
ClickHouse | ClickHouse | 539,649,201 | 7,213 | alexey-milovidov | PS.
```
:) SELECT uniqCombined(reinterpretAsString(number)) FROM numbers_mt(10000000000)
SELECT uniqCombined(reinterpretAsString(number))
FROM numbers_mt(10000000000)
┌─uniqCombined(reinterpretAsString(number))─┐
│ 9990159598 │ ... | 2019-10-08T18:45:05 | 2019-10-08T18:45:05 | {} | MEMBER |
ClickHouse | ClickHouse | 538,971,118 | 7,147 | Tasselmi | > select arrayConcat([arr[1]], arrayFilter(x,y -> x = y, arraySlice(arr, 2), arraySlice(arr, 1, -1))) from ( select [1, 2, 2, 3, 3, 2, 2] arr )
>
> SELECT arrayConcat([arr[1]], arrayFilter(x,y -> y!=0, arr ,arrayDifference(arr))) from (select [1, 2, 2, 3, 3, 2] arr)
The first method is more universal. It tak... | 2019-10-07T11:48:48 | 2019-10-07T11:48:48 | {} | NONE |
ClickHouse | ClickHouse | 539,185,190 | 7,213 | azat | >this looks odd to me that uint64 only for string but this is another story
Actually looking into this more and more, and I cannot verify that this will change in-memory/on-disk representation, am I missing something? | 2019-10-07T20:15:54 | 2019-10-07T20:15:54 | {} | MEMBER |
ClickHouse | ClickHouse | 539,218,159 | 7,217 | isaachui | Thanks! That seemed to have worked! | 2019-10-07T21:45:50 | 2019-10-07T21:45:50 | {} | NONE |
ClickHouse | ClickHouse | 539,237,604 | 7,207 | den-crane | RSS = 1752780 - 1.7 GB
It's totally fine. Because a mark cache = 5BG. (check with `system drop mark cache;`)
Probably it will stabilize at 15GB.
If the system has <32GB Ram it needs a specials configuration.
| 2019-10-07T22:51:19 | 2019-10-07T22:52:37 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,459,397 | 7,214 | 4ertus2 | Related #4567
We need to detect ambiguous columns. It's possible to check table aliases at this phase. | 2019-10-08T10:51:56 | 2019-10-08T10:51:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,676,744 | 6,802 | alexey-milovidov | @Akazz This is still to do. And only a few steps remain. | 2019-10-08T19:57:14 | 2019-10-08T19:57:14 | {} | MEMBER |
ClickHouse | ClickHouse | 539,397,123 | 7,104 | blinkov | merged in #7220 | 2019-10-08T07:59:11 | 2019-10-08T07:59:11 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,495,139 | 7,224 | PaulAnnekov | @tavplubix Again, not so easy to use and maintain:
- need to enable it first
- need to deal with clearing old data (`The storage period for logs is unlimited. Logs aren't automatically deleted from the table. You need to organize the removal of outdated logs yourself.`)
- need to integrate with https://github.com/f... | 2019-10-08T12:44:31 | 2019-10-08T12:44:56 | {} | NONE |
ClickHouse | ClickHouse | 539,669,197 | 7,221 | alexey-milovidov | How to validate:
```
SET max_memory_usage = 100000000;
SELECT intDiv(number, 16384) AS k, uniqCombined(number % 16384) FROM numbers(16384000) GROUP BY k;
```
Each state is using no more than 96KiB.
Please add a test and check whether it fails after these modifications. | 2019-10-08T19:37:44 | 2019-10-08T19:38:07 | {} | MEMBER |
ClickHouse | ClickHouse | 539,683,110 | 7,221 | alexey-milovidov | > Also I checked this in debugger, and got this:
1<<13 with String: getBufferSizeInBytes() == 131072
1<<14 with UInt32: getBufferSizeInBytes() == 262144
That said that even for String hash table can be bigger then HLL
The hash table is resized (or converted to HLL) when it is reached next power of two (if... | 2019-10-08T20:13:00 | 2019-10-08T20:15:37 | {} | MEMBER |
ClickHouse | ClickHouse | 539,163,889 | 7,113 | alexey-milovidov | > because the NULL value shouldn't get past the ifNull function.
It works as expected. For minmax of partition expression, values of source columns are tracked, not the value of full expression. | 2019-10-07T19:19:39 | 2019-10-07T19:19:39 | {} | MEMBER |
ClickHouse | ClickHouse | 539,499,046 | 7,144 | den-crane | > I'd try something like ../../operations/access_rights.md
Thank you. I had the same idea. | 2019-10-08T12:54:30 | 2019-10-08T12:54:30 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,516,502 | 7,207 | den-crane | As I can guess you have 16GB. It's hardly viable for production
You can reduce `mark_cache_size` to 2GB but it's not a hard limit it's a goal.
max_bytes_before_external_group_by = 5GB
max_bytes_before_external_sort = 5GB
https://clickhouse.yandex/docs/en/query_language/select/#select-group-by-in-external-memory
... | 2019-10-08T13:35:41 | 2019-10-08T13:35:41 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,676,101 | 7,232 | den-crane | duplicate for https://github.com/ClickHouse/ClickHouse/issues/7194
Fixed here https://github.com/ClickHouse/ClickHouse/pull/7155/files#diff-bcf50b560562d3f0464f75823443402eR718 | 2019-10-08T19:55:38 | 2019-10-08T19:56:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,683,807 | 7,221 | alexey-milovidov | > Hm, but don't we need care only about on-disk representation? (on disk it should not greater then 65546 bytes -- (1<<14)*4+8+2)
We care about memory size first. The size on disk (and the size of serialized state transferred over network; and how compressible it will be) also matters. But the sizes in memory must b... | 2019-10-08T20:14:48 | 2019-10-08T20:14:48 | {} | MEMBER |
ClickHouse | ClickHouse | 539,684,159 | 7,221 | azat | >resized when it is reached next power of two
Indeed
>It should be of size 64K for 2^13 - 1 elements.
So, the in-memory representation is also significant, and the size of the hash table should be reduced by one, instead of increased by the power of two? | 2019-10-08T20:15:43 | 2019-10-08T20:15:43 | {} | MEMBER |
ClickHouse | ClickHouse | 539,686,009 | 7,221 | alexey-milovidov | > Hm, but don't we need care only about on-disk representation? (on disk it should not greater then 65546 bytes -- (1<<14)*4+8+2)
The size is motivated by memory requirements.
On disk representation is a different story (it's less important scenario - AggregatingMergeTrees and external aggregation). BTW, data on di... | 2019-10-08T20:20:11 | 2019-10-08T20:20:11 | {} | MEMBER |
ClickHouse | ClickHouse | 538,963,463 | 6,742 | akuzm | I'm not woking on this now, so I'll remove myself from the assignees list. | 2019-10-07T11:29:21 | 2019-10-07T11:29:21 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,106,528 | 7,207 | akuzm | Do you use any dictionaries? We fixed a rare memory leak with them recently, but 19.9 doesn't have the fix: https://github.com/ClickHouse/ClickHouse/pull/6447 | 2019-10-07T16:55:41 | 2019-10-07T16:55:41 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,166,629 | 7,215 | akuzm | It would be interesting to check the coverage information for this PR, so that we can be sure that cases are covered by the tests. Parsing is notoriously hard to get right. | 2019-10-07T19:27:17 | 2019-10-07T19:27:17 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,177,850 | 469 | tavplubix | Resolved in #5625 | 2019-10-07T19:57:10 | 2019-10-07T19:57:10 | {} | MEMBER |
ClickHouse | ClickHouse | 539,214,026 | 7,188 | filimonov | you can try also
```
select count() from table xyz final;
``` | 2019-10-07T21:33:48 | 2019-10-07T21:33:48 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,365,622 | 7,213 | azat | > Hash table.
Interesting, since there is 32-bit hash for non-string types, hash table (for medium cardinality) can be expanded by the power of 2 for non-string types, and this will make the value exact for up to 2^14 elements.
> You can radically change uniqCombined function to use 64-bit hash everywhere
This... | 2019-10-08T06:45:30 | 2019-10-08T06:45:30 | {} | MEMBER |
ClickHouse | ClickHouse | 539,419,527 | 7,223 | blinkov | It probably has something to do with background process merging table into less files. Could you show either CREATE TABLE and SELECT queries? | 2019-10-08T08:56:22 | 2019-10-08T08:56:22 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,494,244 | 7,144 | blinkov | > I don't see a problem with Check. Maybe it's a false positive. Please merge it. I'll fix it later if real issue after rendering into HTML.
@den-crane this might be a bug in docs build scripts, but if we just merge there will be no rendered html to check, it'll just stop releasing new versions until this is fixed | 2019-10-08T12:42:23 | 2019-10-08T12:42:23 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 539,535,601 | 7,230 | alex-zaitsev | Do you know of arrayEnumerate() function? It generates indices. Also note that array indices in ClickHouse are 1-based, so if you use indices later to access the same array, your approach does not work. So better to use:
```SELECT arrayMap((x, y) -> (x, y), arrayEnumerate(['a', 'b', 'c']), ['a', 'b', 'c'])```
Doe... | 2019-10-08T14:17:56 | 2019-10-08T14:17:56 | {} | CONTRIBUTOR |
ClickHouse | ClickHouse | 538,915,496 | 7,155 | alesapin | Manually cherry-picked into 19.15. | 2019-10-07T09:26:27 | 2019-10-07T09:26:27 | {} | MEMBER |
ClickHouse | ClickHouse | 539,240,748 | 7,187 | lorenz | It's not about that specific optimization (as I've said above) but about ClickHouse reading whole columns when it only needs a tiny amount of values. The workaround is application-specific but the solution would actually be generalizable. | 2019-10-07T23:02:44 | 2019-10-07T23:02:44 | {} | NONE |
ClickHouse | ClickHouse | 539,371,355 | 7,207 | ishirav | @akuzm - no, we don't use dictionaries.
@den-crane - looks like you are right, after dropping the mark cache the RSS went down to about 800MB. What kind of configuration are you suggesting to use, as a function of the system RAM? I see that it's possible to configure the mark_cache_size, but the docs claim that 5GB is... | 2019-10-08T07:01:44 | 2019-10-08T07:01:44 | {} | NONE |
ClickHouse | ClickHouse | 539,459,177 | 7,223 | MahmoudGoda0 | Hi @blinkov
Do you have any idea why i have diff part numbers for the 3 tables ? | 2019-10-08T10:51:10 | 2019-10-08T10:51:10 | {} | NONE |
ClickHouse | ClickHouse | 539,421,396 | 7,223 | MahmoudGoda0 | It seems you are right, i just realized the below results
```
SELECT count(partition)
FROM system.parts
WHERE table = 'table1'
```
results = 75
```
SELECT count(partition)
FROM system.parts
WHERE table = 'table2'
```
results = 74
```
SELECT count(partition)
FROM system.parts
WHERE table = 'table3'
```
... | 2019-10-08T09:01:07 | 2019-10-15T13:07:52 | {} | NONE |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.