Support string repeat SQL

[ttnghia]

This PR implements repeat SQL, and closes #68.

Example:

strs = ['abc', null, '',  'xyz']
out  = repeat(strs, 3)
out is ['abcabcabc', null, '',  'xyzxyzxyz']

repeatTimes = [1, 2, 3, 4]
out = repeat(strs, repeatTimes)
out is ['abc', null, '',  'xyzxyzxyzxyz']

[jlowe]

@ttnghia the commit needs to be signed off, see the signoff section in the contributing guide for details.

[sameerz]

@ttnghia please retarget to branch 21.10 when you get a chance.

[jlowe]

Looks good to me other than a small assert result nit.

[jlowe]

build

[revans2]

The latest test run failed because of a breaking change in the cudf APIs that was merged in and the corresponding plugin change has not gone in yet.

[revans2]

build

[ttnghia]

build

[pxLi]

hi. please check the build log before trigger again thx!

[2021-08-04T16:34:40.039Z] found modified files during mvn verify:
[2021-08-04T16:34:40.039Z]  M docs/configs.md
[2021-08-04T16:34:40.039Z]  M docs/supported_ops.md
[2021-08-04T16:34:40.039Z] [ERROR] Command execution failed.

this PR did not include RapidsConf and SupportedOpsDocs update well, so it failed the check

[ttnghia]

hi. please check the build log before trigger again thx!

[2021-08-04T16:34:40.039Z] found modified files during mvn verify:
[2021-08-04T16:34:40.039Z]  M docs/configs.md
[2021-08-04T16:34:40.039Z]  M docs/supported_ops.md
[2021-08-04T16:34:40.039Z] [ERROR] Command execution failed.

this PR did not include RapidsConf and SupportedOpsDocs update, so it failed the check

Oh thanks for letting me know. I have updated the docs.
@jlowe I hate to ask you to approve the third time but I have to. I also hate github that discards PR approval every time I push. Maybe we need to modify that behavior for better convenience.

[pxLi]

hi. please check the build log before trigger again thx!

[2021-08-04T16:34:40.039Z] found modified files during mvn verify:
[2021-08-04T16:34:40.039Z]  M docs/configs.md
[2021-08-04T16:34:40.039Z]  M docs/supported_ops.md
[2021-08-04T16:34:40.039Z] [ERROR] Command execution failed.

this PR did not include RapidsConf and SupportedOpsDocs update, so it failed the check

Oh thanks for letting me know. I have updated the docs.
@jlowe I hate to ask you to approve the third time but I have to. I also hate github that discards PR approval every time I push. Maybe we need to modify that behavior for better convenience.

Actually I think we should not, otherwise people could push new commits which could do whatever they want to our repo after the initial approval, this could be horrible :P

[ttnghia]

Oh thanks for letting me know. I have updated the docs.
@jlowe I hate to ask you to approve the third time but I have to. I also hate github that discards PR approval every time I push. Maybe we need to modify that behavior for better convenience.

Actually I think we should not, otherwise people could push new commits which could do whatever they want to our repo after the initial approval, this could be horrible :P

People should know what they are doing, not "whatever they want" like that 😃. I observed (from cudf) that after approvals the authors usually push more minor changes related to docs/typos etc. If every new change requires approvals again then it is very inconvenient for both the authors and the reviewers.

Of course, if there are new changes in the code then the authors can request re-review anytime.

[pxLi]

Oh thanks for letting me know. I have updated the docs.
@jlowe I hate to ask you to approve the third time but I have to. I also hate github that discards PR approval every time I push. Maybe we need to modify that behavior for better convenience.

Actually I think we should not, otherwise people could push new commits which could do whatever they want to our repo after the initial approval, this could be horrible :P

People should know what they are doing, not "whatever they want" like that 😃. I observed (from cudf) that after approvals the authors usually push more minor changes related to docs/typos etc. If every new change requires approvals again then it is very inconvenient for both the authors and the reviewers.

Of course, if there are new changes in the code then the authors can request re-review anytime.

haha yeah, I mean cudf approval is horrible. For our repo, if PR tests pass pre-merge CI but requires to add some minor changes like comments/docs only (not for generated docs like configs.md or supported_ops.md), people could just add [skip ci] to the title to skip the additional full tests run w/ new commit approval :)

[sameerz]

Oh thanks for letting me know. I have updated the docs.
@jlowe I hate to ask you to approve the third time but I have to. I also hate github that discards PR approval every time I push. Maybe we need to modify that behavior for better convenience.

Actually I think we should not, otherwise people could push new commits which could do whatever they want to our repo after the initial approval, this could be horrible :P

People should know what they are doing, not "whatever they want" like that 😃. I observed (from cudf) that after approvals the authors usually push more minor changes related to docs/typos etc. If every new change requires approvals again then it is very inconvenient for both the authors and the reviewers.
Of course, if there are new changes in the code then the authors can request re-review anytime.

haha yeah, I mean cudf approval is horrible. For our repo, if PR tests pass pre-merge CI but requires to add some minor changes like comments/docs only (not for generated docs like configs.md or supported_ops.md), people could just add [skip ci] to the title to skip the additional full tests run w/ new commit approval :)

It may be inconvenient, but a review on a change, even a small one, is worth the trouble to minimize potential problems. Doing a review on every change is a good habit. We can still move quickly while doing the reviews.

[jlowe]

build