time-series

Hi,

first thanks for this awesome software. But I have some trouble and I don't know how to proceed. First I try to find out what else I could provide for this report to be a good bug report.

First, I use the latest version (v1.14.0) from openSUSE build Service. I know, I should compile it by my self to avoid any other causes for this behavior. But for now I still use this package from there

Feature Request

support an input plugin of the following type:

[[inputs.test]]
  metrics = [
    'weather,state=ny temperature=81.3'
  ]

Proposal:

support the above

Current behavior:

possible through creating a file and using inputs.file or inputs.tail

Desired behavior:

above

Use case: <!-- [Why is this important (helps with prioritizing

it's becoming more time-consuming and error-prone to manually re-test all the demos following internal refactorings and API adjustments.

now that the API is fleshed out a bit, it's possible to test a large amount of code (non-granularly) without having to simulate all interactions via Puppeteer or similar.

a lot of code can already be regression-tested by simply running all the demos and val

Scenario

create table tab(f float);
insert into tab(f) values(4273.8);

Results in: org.postgresql.util.PSQLException: ERROR: inconvertible types: DOUBLE -> FLOAT [from=4273.8, to=f]

Expectation

Possibly enhance type compatibility check (SqlCompiler.isAssignableFrom) with extra cases for numeric types:

• Float <- Double: if the double fits in the Float, then

Is your feature request related to a problem? Please describe.
The current dataset with exog data for forecasting is load_longley() but it has only len(y) of 16, so e.g. AutoARIMA fails with this when doing a default train_test_split. Also for vector forecasting, which is probably coming soon, we might need a longer dataset for some forecasters.

**Describe the solution you'd lik

提问题？
请把问题现象、相关日志、初步排查结果写清楚，即提供尽量多的线索，这样方便我们定位问题。

提建议？
请写明白建议原因，即为何有此建议。

提修复PR？
请说明为何做此修改，需要做好测试。

提Feature？
建议先将想法提出，不着急写代码，与社区同道沟通，大家觉得OK再做代码实现

---

另外，我们是在做开源软件，不是服务于您的企业，做您的乙方，您一分钱没付，就不要颐指气使了，认同我们这个社区，则参与进来，不认同，就不参与，谢谢理解

Use case:

1.) A user may want to backup all tables but no metadata like users, privileges, etc. without explicitly defining each table inside the CREATE SNAPSHOT statement.

2.) A user may want to transfer users & privileges, custom analyzers or user-defined-functions from one cluster to another without backing up a complete cluster including all data (tables).

*Feature description

Description

(A clear and concise description of what the feature is.)

• util.cumsum implementation https://github.com/awslabs/gluon-ts/blob/master/src/gluonts/mx/util.py#L326 does not scale under mx.ndarray
• cumsum is 2-5 times slower than nd.cumsum under both mx.sym and mx.ndarray, and even fails for large 4-dim input

Sample test

Code

# import ...
def test_

Hi, Thanks for the awesome library!

So I am running a Kmeans on lots of different datasets, which all have roughly four shapes, so I initialize with those shapes and it works well, except for just a few times. There are a few datasets that look different enough that I end up with empty clusters and the algorithm just hangs ("Resumed because of empty cluster" again and again).

I conceptually