3 ms·
I totally agree. Complete pile of crap. The search is quite useless (try searching for something as simple as all tickets to which you added a comment in the la
by gotofritz 10y ago
I totally agree. Complete pile of crap. The search is quite useless (try searching for something as simple as all tickets to which you added a comment in the last week...no chance). Unless you get a plugin, but who writes these plugins? Can you trust them? Who's assessing them? And how much do they cost? And what if you haven't got admin rights? Keeping track of stuff should be the main functionality of the product, it should be slick out of the box. Not to mention all the other stuff, useless kanban boards, bulk operation, and all the rest of it. I find myself clicking for whole minutes to complete simple tasks.
I think Atlassian have too many pots on the fire and they can't keep up with it.
- locopati 10y agoInstall the Script Runner Plugin: issueFunction in commented("by username after -7d")
- mikey_p 10y ago> try searching for something as simple as all tickets to which you added a comment in the last week...no chance That's actually a crazy complicated query if you think about. Not so much from a straight sequel side, but how exactly would you build a friendly UI that lets you specify the conditions that would give you the resulting query? Everyone here is bashing on JIRA UI for being too complicated and trying to do to much, but then you have people wanting features like this.
- wruza 10y agoIf it is too complicated but can't do this, then maybe it is complicated in a wrong way. Second, "tickets to which you added a comment in the last week" in other words are "active tickets for a week" or more shortly "tickets related to me", which seems to be pretty useful for, e.g., a welcome screen for a developer/manager/whatever. One doesn't look at new tickets each 15 minutes, idk why all darned trackers show that after login.
- gotofritz 10y agoWell I have no idea how they store their data, but if it was straight T-SQL and assuming a table for tickets, a table for comments, and a table linking the two (typical way to implemente 1-to-M relationships) something like (pseudo SQL) SELECT * FROM tickets JOIN ticketsToComments on tickets.ID = ticketsToComments.ticket_ID join comments on ticketsToComments.comment_ID = comments.ID JOIN autor on comments.autor_ID = autor.ID where autor_ID = me and tickets.date < 1 week It's not rocket science
- dragonwriter 10y ago> Well I have no idea how they store their data, but if it was straight T-SQL and assuming a table for tickets, a table for comments, and a table linking the two (typical way to implemente 1-to-M relationships) Wouldn't 1:M just be a typical FK relationship, a join table is only needed for either M:N relationships or relationships that have additional attributes belonging to the relationship rather than either of the joined tables.
- gotofritz 10y agoYes you are right, it was late at night :-) That makes it even simpler, removing one join... SELECT * FROM tickets JOIN comments on comments.ticket_ID = ticket.ID JOIN autor on comments.autor_ID = autor.ID where autor_ID = me and tickets.date < 1 week