Thanks Bryan. Glad to see this is on the radar. I would recommend maybe an incremental approach. First, implement the simpler lock/unlock mechanism as phase 1. Phase 2 could then be the full-blown undo which is certainly a magnitude order more complex to implement. At least some basic safety could be introduced earlier in a phased approach.
On Sat, Oct 31, 2015 at 10:00 AM, Bryan Bende <[email protected]> wrote: > We previously had a very long thread about this topic [1] which I think > resulted in the idea of "undo" [2]. > > [1] > https://mail-archives.apache.org/mod_mbox/nifi-dev/201508.mbox/%3ccaljk9a6nn5y7ep7-dw5yeb-33m7zjvvw+h2nj4auccmlpeh...@mail.gmail.com%3E > > [2] > https://cwiki.apache.org/confluence/display/NIFI/Configuration+Management+of+Flows > > On Sat, Oct 31, 2015 at 9:46 AM, Mark Payne <[email protected]> wrote: >> >> Matt, >> >> Just as an FYI, this was changed recently (perhaps 0.3.0?) so that you can >> delete a process group that is not empty. >> >> However, it does have to have all components stopped, no incoming >> connections, and all connections must be empty. >> >> Thanks >> -Mark >> >> >> On Oct 31, 2015, at 3:05 AM, Matthew Clarke <[email protected]> >> wrote: >> >> You cannot delete a process group unless it is empty, so you don't need to >> worry about deleting nested sub process groups. I myself see the value in a >> 'are you sure' pop-up. It just needs to be implemented in a smart way. I >> created a ticket -- Nifi-1089 >> >> On Oct 31, 2015 2:20 AM, "Mark Petronic" <[email protected]> wrote: >>> >>> Thanks. Ya, I realized the checks in place for deletes under the >>> conditions you listed. And, I can understand the 'annoying' factor, too. But >>> maybe the DFM could default to a popup "Are you sure" with a "[ ] Don't ask >>> me again" check box option. Most applications either ask or have undo. This >>> has neither so one must be careful or you could accidentally wack a heck of >>> a lot of graphs (especially with nested sub processor groups) with one >>> accidental hit of the delete key. >>> >>> On Sat, Oct 31, 2015 at 1:25 AM, Matthew Clarke >>> <[email protected]> wrote: >>>> >>>> There are check to make sure you can't delete a processor if it is >>>> running or has FlowFiles in any of its connections. There are checks to >>>> make >>>> sure that connections can not be deleted it they contain FlowFiles also. >>>> You >>>> are correct that if none of these conditions exist, there is no 'are you >>>> sure' second step. I don't see why this can't be done, but it could also >>>> get annoying for users who are constantly changing their graph. >>>> >>>> There is also no "undo" feature. NiFi does not lock the graph to a >>>> user. At anytime another user can access the graph and make changes at the >>>> same time. Whomever hits OK or apply first wins. The other user will lose >>>> any changes they were in the process of making, so coordination between >>>> user >>>> is crucial. Since multiple use can make changes to the same component, how >>>> do you undo your change without possibly undoing someone else's change. >>>> >>>> The only possibility I see is a very limited undo feature. One >>>> that will allow you to undo only the last change made. So if someone else >>>> makes a change after you, you will no longer be able to undo your change. >>>> >>>> On Oct 30, 2015 11:48 PM, "Mark Petronic" <[email protected]> >>>> wrote: >>>>> >>>>> With a simple inadvertent key press, you can easily delete a processor >>>>> or a whole processor group or set of selected components without even an >>>>> "Are you sure?" confirmation and I see no way to 'undo' the mistake. IMO, >>>>> deleting should be a two-step process given the amount of effort one might >>>>> make building a complex flow. Or, am I missing something here? >>> >>> >> >
