"Replace" button for files in draft

24 views
Skip to first unread message

Sebastian Karcher

unread,
Apr 16, 2018, 2:35:53 PM4/16/18
to Dataverse Users Community
Is there any reason the "Replace" http://guides.dataverse.org/en/latest/user/dataset-management.html#replace-files functionality for files isn't available for datasets in Draft?
That's actually where I'd like to use it the most: I've uploaded the wrong file but want to keep my description & tag. This would also be super useful for curation. Before opening a ticket, I wanted to see if there were perhaps reasons against this that I'm missing or if its absence is actually just a bug/oversight (tested both in our 4.8.6 and at demo.dataverse.org.

Thanks!
Sebastian

Philip Durbin

unread,
Apr 16, 2018, 4:35:36 PM4/16/18
to dataverse...@googlegroups.com
Hi Sebastian,

Someone else might be able to weigh in with a better recollection of why draft files cannot be replaced, but we may have been thinking that for a file to be replaced in the Dataverse sense we need to persist the "previousdatafileid" in the "datafile" table[1] and when a file is in draft there is no previous file id yet.

With regard to your user story of wanting to be able to keep your description and tag, I'm afraid I have some bad news for you. Please see https://github.com/IQSS/dataverse/issues/4380 for the following feedback:

"When replacing a file the original file's metadata doesn't not carryover. I would expect the files description and tags to be the same as the original file since I'm replacing the file. There is also no warning that the metadata and tags will have to be reentered."

The feedback you're giving is valuable so please feel free to comment on the issue above or open new issues.

Thanks,



--
You received this message because you are subscribed to the Google Groups "Dataverse Users Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataverse-community+unsub...@googlegroups.com.
To post to this group, send email to dataverse-community@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dataverse-community/8ba1af45-78b1-42ca-8c61-976a08507e4d%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.



--

Sebastian Karcher

unread,
Apr 16, 2018, 10:55:02 PM4/16/18
to dataverse...@googlegroups.com
Thanks Phil,
I'm glad I asked - I've added our user story to that ticket. Before that's closed, I don't think allowing replace during draft makes any sense (and could in fact lead to unexpected [meta]data loss).
Sebastian

On Mon, Apr 16, 2018 at 4:35 PM, Philip Durbin <philip...@harvard.edu> wrote:
Hi Sebastian,

Someone else might be able to weigh in with a better recollection of why draft files cannot be replaced, but we may have been thinking that for a file to be replaced in the Dataverse sense we need to persist the "previousdatafileid" in the "datafile" table[1] and when a file is in draft there is no previous file id yet.

With regard to your user story of wanting to be able to keep your description and tag, I'm afraid I have some bad news for you. Please see https://github.com/IQSS/dataverse/issues/4380 for the following feedback:

"When replacing a file the original file's metadata doesn't not carryover. I would expect the files description and tags to be the same as the original file since I'm replacing the file. There is also no warning that the metadata and tags will have to be reentered."

The feedback you're giving is valuable so please feel free to comment on the issue above or open new issues.

Thanks,


On Mon, Apr 16, 2018 at 2:35 PM, Sebastian Karcher <sebastiankarcher2008@u.northwestern.edu> wrote:
Is there any reason the "Replace" http://guides.dataverse.org/en/latest/user/dataset-management.html#replace-files functionality for files isn't available for datasets in Draft?
That's actually where I'd like to use it the most: I've uploaded the wrong file but want to keep my description & tag. This would also be super useful for curation. Before opening a ticket, I wanted to see if there were perhaps reasons against this that I'm missing or if its absence is actually just a bug/oversight (tested both in our 4.8.6 and at demo.dataverse.org.

Thanks!
Sebastian

--
You received this message because you are subscribed to the Google Groups "Dataverse Users Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataverse-community+unsubscribe...@googlegroups.com.
--

--
You received this message because you are subscribed to the Google Groups "Dataverse Users Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataverse-community+unsub...@googlegroups.com.
To post to this group, send email to dataverse-community@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dataverse-community/CABbxx8G9Y-H6PMLaoOjwcVGLrLXpeKJd%2BreO75OcYBJxckfg-Q%40mail.gmail.com.

For more options, visit https://groups.google.com/d/optout.



--
Sebastian Karcher, PhD
www.sebastiankarcher.com

danny...@g.harvard.edu

unread,
Apr 17, 2018, 7:31:53 AM4/17/18
to Dataverse Users Community
Hi,

Thanks for leaving information about the additional scenario on the ticket. The file replace feature was originally added in support of the Prov CPL integration (https://github.com/ProvTools/prov-cpl/), in which "replacement" is a specific relationship between two files in different dataset versions. For example, file x replaced file y and I'm adding provenance information as to why. I don't think we'll revisit this functionality in the near future, but we can perhaps revisit the documentation. 

- Danny

On Monday, April 16, 2018 at 10:55:02 PM UTC-4, Sebastian Karcher wrote:
Thanks Phil,
I'm glad I asked - I've added our user story to that ticket. Before that's closed, I don't think allowing replace during draft makes any sense (and could in fact lead to unexpected [meta]data loss).
Sebastian
On Mon, Apr 16, 2018 at 4:35 PM, Philip Durbin <philip...@harvard.edu> wrote:
Hi Sebastian,

Someone else might be able to weigh in with a better recollection of why draft files cannot be replaced, but we may have been thinking that for a file to be replaced in the Dataverse sense we need to persist the "previousdatafileid" in the "datafile" table[1] and when a file is in draft there is no previous file id yet.

With regard to your user story of wanting to be able to keep your description and tag, I'm afraid I have some bad news for you. Please see https://github.com/IQSS/dataverse/issues/4380 for the following feedback:

"When replacing a file the original file's metadata doesn't not carryover. I would expect the files description and tags to be the same as the original file since I'm replacing the file. There is also no warning that the metadata and tags will have to be reentered."

The feedback you're giving is valuable so please feel free to comment on the issue above or open new issues.

Thanks,


On Mon, Apr 16, 2018 at 2:35 PM, Sebastian Karcher <sebastiankarcher2008@u.northwestern.edu> wrote:
Is there any reason the "Replace" http://guides.dataverse.org/en/latest/user/dataset-management.html#replace-files functionality for files isn't available for datasets in Draft?
That's actually where I'd like to use it the most: I've uploaded the wrong file but want to keep my description & tag. This would also be super useful for curation. Before opening a ticket, I wanted to see if there were perhaps reasons against this that I'm missing or if its absence is actually just a bug/oversight (tested both in our 4.8.6 and at demo.dataverse.org.

Thanks!
Sebastian

--
You received this message because you are subscribed to the Google Groups "Dataverse Users Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataverse-community+unsub...@googlegroups.com.
To post to this group, send email to dataverse-community@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dataverse-community/8ba1af45-78b1-42ca-8c61-976a08507e4d%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "Dataverse Users Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataverse-community+unsub...@googlegroups.com.
To post to this group, send email to dataverse-community@googlegroups.com.
 To view this discussion on the web visit https://groups.google.com/d/msgid/dataverse-community/CABbxx8G9Y-H6PMLaoOjwcVGLrLXpeKJd%2BreO75OcYBJxckfg-Q%40mail.gmail.com.

For more options, visit https://groups.google.com/d/optout.

Philip Durbin

unread,
Apr 17, 2018, 8:08:23 AM4/17/18
to dataverse...@googlegroups.com
It's certainly inconvenient that using the file replace feature means you have to re-enter the description of the file and re-enter tags, but I wouldn't describe this as metadata loss. File replace creates a new draft version of the dataset so you can always refer back to the previously published version to see the description, etc.

To view this discussion on the web visit https://groups.google.com/d/msgid/dataverse-community/CABbxx8G9Y-H6PMLaoOjwcVGLrLXpeKJd%2BreO75OcYBJxckfg-Q%40mail.gmail.com.

For more options, visit https://groups.google.com/d/optout.



--
Sebastian Karcher, PhD
www.sebastiankarcher.com

--
You received this message because you are subscribed to the Google Groups "Dataverse Users Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataverse-community+unsub...@googlegroups.com.
To post to this group, send email to dataverse-community@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dataverse-community/CAOSYSD7GxqwdNVDgDN5Ft6J%2BP1n0Nd9Dz9ALtFBKzKmeA7MYtA%40mail.gmail.com.

For more options, visit https://groups.google.com/d/optout.

Sebastian Karcher

unread,
Apr 17, 2018, 8:43:36 AM4/17/18
to Dataverse Users Community
Phil: right, in its current version there's no such risk. I was saying that allowing file replace on draft -- where there's no older version of the dataset that one can go back to -- without preserving the metadata would add the potential for metadata loss, in other words, addressing #4380 would be a precondition for what I'm asking about here. 

Danny: thanks for the background on this, that makes sense and would indeed be helpful to document. I think two sentences (...will only be available once dataset is published ... is intended for provenance...) would be all it takes to clarify this. Re: no revisit the functionality, does that mean you also wouldn't take PRs (not committing to anything...) or just that you're not going to work on this at IQSS?

Thanks both!
Sebastian


On Tuesday, April 17, 2018 at 8:08:23 AM UTC-4, Philip Durbin wrote:
It's certainly inconvenient that using the file replace feature means you have to re-enter the description of the file and re-enter tags, but I wouldn't describe this as metadata loss. File replace creates a new draft version of the dataset so you can always refer back to the previously published version to see the description, etc.
On Mon, Apr 16, 2018 at 10:54 PM, Sebastian Karcher <kar...@u.northwestern.edu> wrote:
Thanks Phil,
I'm glad I asked - I've added our user story to that ticket. Before that's closed, I don't think allowing replace during draft makes any sense (and could in fact lead to unexpected [meta]data loss).
Sebastian
On Mon, Apr 16, 2018 at 4:35 PM, Philip Durbin <philip...@harvard.edu> wrote:
Hi Sebastian,

Someone else might be able to weigh in with a better recollection of why draft files cannot be replaced, but we may have been thinking that for a file to be replaced in the Dataverse sense we need to persist the "previousdatafileid" in the "datafile" table[1] and when a file is in draft there is no previous file id yet.

With regard to your user story of wanting to be able to keep your description and tag, I'm afraid I have some bad news for you. Please see https://github.com/IQSS/dataverse/issues/4380 for the following feedback:

"When replacing a file the original file's metadata doesn't not carryover. I would expect the files description and tags to be the same as the original file since I'm replacing the file. There is also no warning that the metadata and tags will have to be reentered."

The feedback you're giving is valuable so please feel free to comment on the issue above or open new issues.

Thanks,


On Mon, Apr 16, 2018 at 2:35 PM, Sebastian Karcher <sebastiank...@u.northwestern.edu> wrote:
Is there any reason the "Replace" http://guides.dataverse.org/en/latest/user/dataset-management.html#replace-files functionality for files isn't available for datasets in Draft?
That's actually where I'd like to use it the most: I've uploaded the wrong file but want to keep my description & tag. This would also be super useful for curation. Before opening a ticket, I wanted to see if there were perhaps reasons against this that I'm missing or if its absence is actually just a bug/oversight (tested both in our 4.8.6 and at demo.dataverse.org.

Thanks!
Sebastian

--
You received this message because you are subscribed to the Google Groups "Dataverse Users Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataverse-community+unsub...@googlegroups.com.
To post to this group, send email to dataverse...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Dataverse Users Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataverse-community+unsub...@googlegroups.com.
To post to this group, send email to dataverse...@googlegroups.com.



--
Sebastian Karcher, PhD
www.sebastiankarcher.com

--
You received this message because you are subscribed to the Google Groups "Dataverse Users Community" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dataverse-community+unsub...@googlegroups.com.
To post to this group, send email to dataverse...@googlegroups.com.

danny...@g.harvard.edu

unread,
Apr 17, 2018, 11:53:59 AM4/17/18
to Dataverse Users Community
Hey Sebastian, we're willing to take PRs! If you or any other community members want to take this on we should probably discuss first. We could discuss solving the problem in a different way or discuss an implementation that doesn't later block provenance work. Thanks for asking!

Derek Murphy

unread,
Apr 17, 2018, 12:51:19 PM4/17/18
to Dataverse Users Community
Hi Sebastian,

Thank you for reporting on your encounter with file replace. I'll make some edits to the documentation to help clarify the points of confusion you saw.


On Tuesday, April 17, 2018 at 8:43:36 AM UTC-4, Sebastian Karcher wrote:
Reply all
Reply to author
Forward
0 new messages