Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
    Patrick McSweeney
    @patrickmcsweeney
    we could make the default a 100 and it would still be a significant improvement.
    the alternative could be that we just an appropriately sized maxlength on all the default fields in eprints_fields.pl
    Lizz Jennings
    @icklecows
    100 would cover most proper uses of namedsets I think. I mean people shouldn't be writing sentences!
    But URIs are a valid way of doing it, and they can get a bit lengthy
    Patrick McSweeney
    @patrickmcsweeney
    Yeah and linked data URIs can get a bit out of control depending on what your named set is about
    The issue i have now is that ive been round eprints.soton putting limits on all my set fields
    but i have to work out a slick way to generate script which generates alter statements to change my database to fit.
    ive got dev pre-prod and prod so there is plenty of sandpits to play in
    Lizz Jennings
    @icklecows
    I suppose it really depends if it's a housekeeping issue at soton - then you could probably get something to work out what the maximum length of your sets are, and set that as a max. And there are some fields that could almost certainly have 20 or less and no-one would notice
    But if it's global, then I think it needs to be a bit more lenient
    jesusbagpuss
    @jesusbagpuss
    What happens when you try and import (e.g not create via the GUI) some data with over-long data in fields?
    Lizz Jennings
    @icklecows
    For sets you'd still need to have a consistent name, or you get errors anyway...
    Patrick McSweeney
    @patrickmcsweeney
    I havent checked but i suspecit if you make an item in a named set with a key which is too long for the field very bad things happen
    like everything will start up happily and looks like it works
    and when you come to submit to the workflow you get a 500 and all the data from that form stage gets blow away...
    that bug probably exists already but no one has ever written a test to see what happens if you make key which is over 255 long.
    there should probably be some proper validation for that and a graceful failure case....
    Patrick McSweeney
    @patrickmcsweeney
    What should probably happen is that on instatiation the field should validate that all the keys fit in the hole and if they dont it should fail gracefully
    for added brownie points it could do the same check to decide an approrpiately sized varchar
    hmm it appears gitter is shit
    it keeps dropping my messages
    the downside of that approach is adding an item to named set will sometimes involve calling "epadmin update" which maybe undesirable
    jesusbagpuss
    @jesusbagpuss
    storing a value in a namedset doesn't require an epadmin/update - it doesn't look good if there isn't a phrase defined for it - and you have to add it to the 'options' of the field in the workflow - which isn't good.
    If you don't add it to the options in the workflow, you can't retain a value in that field = data loss when someone edits the EPrint. I'm currently wrestling this for a licence field that does more sensible things than the current one!
    Patrick McSweeney
    @patrickmcsweeney
    sorry John i dont think i understood what you were getting at there
    jesusbagpuss
    @jesusbagpuss
    Namedset options (in file): a,b,c,d
    Workflow rendering: select: a / b / c / d
    If (via an import / connected system) we set the value to 'e', next time someone edits the item in the GUI, they're presented with the options a/b/c/d - and can't retain the already-set value 'e'
    At least that's what was happening on our 3.3.10 server!
    Patrick McSweeney
    @patrickmcsweeney
    oh really? that had not occurred to me.
    can you not just add e to the set?
    does that mean you can do $eprint->set_value("yourset", "e"); it just lets you do that?
    because if so thats a bit cheesy isnt it
    or are you putting values directly in the database?
    jesusbagpuss
    @jesusbagpuss
    No - not hacking around in the database - importing things (like EPXML).
    It's useful to be able to store values that aren't defined - maybe a flag on a named set - 'constrain_to_known_values' - but by default, if you've stored a value X, it shouldn't get silently thrown away!
    Patrick McSweeney
    @patrickmcsweeney
    Well how about being able to define a seperate set which gets rendered into the form/
    It would need an option in the form which says "Retain current value"
    jesusbagpuss
    @jesusbagpuss
    Yep - this is exactly what I'm thinking for a new 'Licence' type field - have a default set of e.g. v4 CC licenses (plus a few others), but allow someone to store a URI to any licencse. By default it would allow a saved, but unknown value to be retained. An admin function would also allow an easy way to find 'unknown values' - so instead of rendering them as e.g. just a URI, you can add a phrase for the URI - and render it in a more friendly way!
    Patrick McSweeney
    @patrickmcsweeney
    yeah maybe it should be in the admin section "Manage namedsets" . It shows you the values in all of them and checks whether they have phrases assoicated
    Rotatabol
    @Rotatabol
    Hey guys, I'm not sure if you can help me: I want to configure my eprints repo to access the same archive using two domains. I think this should be done with the {aliases} option in .../myrepo/cfg/cfg.d/10core.pl and reloading all documents with generate* in eprints3/bin. But this don't work for me: I can access the repo with both domains, but it is just the same as an redirect in the apache config. All links on the page are hardlinks that point to the first domain so the second domain is not usable. Do you have any suggestions? I already asked at eprints-tech mailing list, but they can't help me. Thank you very much for your help!
    Lizz Jennings
    @icklecows
    @Rotatabol I don't think there's anyone here who isn't also on the eprints-tech mailing list, but perhaps someone will see this and have an idea
    Matthew Kerwin
    @phluid61
    I've seen it there (and here, too, now) but haven't been able to invest much thought into it. In the past we've had a similar-but-different issue, where I wanted to configured our repo to be HTTPS-by-default (but still allow some endpoints to be accessible under cleartext HTTP). I never really resolved that properly.
    zstoynova
    @zstoynova
    @Rotatabol Were you able to solve your issue with configuring your eprints repo to access the same archive using two domains?
    Rotatabol
    @Rotatabol

    Hey @zstoynova, unfortunately, I was not able to solve it. I created a very ugly workaround: two Apache vhost-configs with the same eprints ID and different domains, one of them using mod_substitute to change all links to its domain. The second vhost with the 'false' domain includes this additional lines:

    <Location "/">
    AddOutputFilterByType INFLATE;SUBSTITUTE;DEFLATE text/html text/plain text/cgi
    Substitute "s;http://example.com;http://example2.com;ni"
    </Location>

    while example.com is the domain known to eprints and example2.com is the additional domain name.

    zstoynova
    @zstoynova
    thank you very much
    Patrick McSweeney
    @patrickmcsweeney
    Yes the only other way i can really think to do that is duplicating the archive directories but with a different host. Out of interest what is the purpose of using two domains for the same repo? it just sounds like it would split the google rank
    Rotatabol
    @Rotatabol
    @patrickmcsweeney we do not need a good google ranking, this is a scientific publication server
    with a very small user group. Our library team needs two different domains for cooperation with publishing companies.
    zstoynova
    @zstoynova
    We would like use git for the development of the Eprints archives. therefore we have a production and a test server for the master and develop branch versions of our repository. Actually our production archives are in fact duplicated on the test server and we would like to access them using <archivename>-test ..... we thought that using an alias with no redirection will solve our problem but the as it was alreaydy explaine all links on the page are hardlinks that point to our production domain so our test domain is not usable.
    zstoynova
    @zstoynova
    Hi @Rotatabol we have tested your solution and everything works except user login. It seems that the original domain address is given as an argument "rel_cgipath" to login redirection address and therefore apache configuration with mod_substitute plays no role. Did you have similar problem or not?
    Rotatabol
    @Rotatabol
    @zstoynova Sorry, i forgot to mention it. I've had this problem too, but this is fine for me - I didn't look for a solution. Please message me if you figure this out!