not laying out correctly with blocking tasks

[Scotty M.]

If I add a ticket, no. 2, and it's blocked by ticket no. 1, and I have a start date for ticket no. 1 but none for no. 2, ticket no. 2 should appear by default at the end of ticket no. 1 (start no. 1 + estimated hours, or spent hours if greater).

[Chris Nelson]

Replying to Scotty M.:

If I add a ticket, no. 2, and it's blocked by ticket no. 1, and I have a start date for ticket no. 1 but none for no. 2, ticket no. 2 should appear by default at the end of ticket no. 1 (start no. 1 + estimated hours, or spent hours if greater).

Yep, the plugin doesn't do scheduling. I'm working on it.

[Hans]

It would also be helpful if it worked the other way round, i.e., any tickets that are blockers to ticket 1 are placed before ticket 1 in the timeline.

Right now, if a ticket has no start or end dates, it gets put right before the milestone's due date, which is good. However, any tickets marked as blockers end up with their end-date set to today, regardless of which is blocking which.

[Chris Nelson]

I hope to post an attempt to fix this in the next week.

[Chris Nelson]

(In [10688]) Refactor task sorting. Refs #9042, #8982.

• Comment sort routine. Minor white space fix.

• "sort" is really "compare".

• Sort all tickets in Gantt the same way we sort children of a parent.

• Don't need tempEnd if we have calc_finish

[Chris Nelson]

(In [10689]) Sort based on dependencies. Refs #9042, #8982.

This nicely puts milestones after their tasks.

This turned up a problem with dependency representation. When we did:

s = '1, 2, 3'
l = s.split(',')

l ended up being ['1', ' 2', ' 3'] so "'2' in l" was false.

So now we strip white space from pred and succ ticket IDs.

Also, Trac milestones don't have successors but an empty list simplifies other code.

[Chris Nelson]

(In [10690]) Schedule tasks by hour, not day. Refs #9042, #8982.

This is an first step toward handling dependencies (so, a four-hour task and an 12-hour task can take just two 8-hour days) but it also addresses a display but that tasks that took a multiple of 8 hours showed up a day too long in the Gantt.

Giving Trac milestones 0 estimates simplifies other code.

[Chris Nelson]

(In [10691]) Add as-late-as-possible scheduling algorithm. Refs #9042, #8982.

[Chris Nelson]

(In [10692]) Add as-soon-as-possible scheduling algorithm. Refs #9042, #8982.

ALAP with all starts changed to finish, successors change to predecessors, etc.

[Chris Nelson]

(In [10693]) Remove dead code (old scheduling algorithm). Refs #9042, #8982.

[Chris Nelson]

(In [10695]) Use adjusted estimate for computing hours, percent complete. Refs #9042, #8982.

• Add configurable default estimate for tasks.

• Add configurable overestimate padding. If work is already more than estimate, we likely need still more time.

• Improve reporting of percent complete. Use projected work, not estimate.

• Remove a stale comment; already fixed.

[anonymous]

This is somewhat experimental and not tested at all for tickets which have start dates specified. Feedback on functionality is welcome. I could use some code reviews, too. My Python is weak and I use recursion and object assignments in ways that may not be optimal.

[Chris Nelson]

(In [10696]) Fix a bug with tickets without due dates. Refs #9042, #8982.

[Chris Nelson]

(In [10710]) Fix several minor errors in commits for scheduling. Refs #9216, #9024, #8982.

Applying a patch from rjollos. Thanks!

[Chris Nelson]

(In [11071]) Refactor PM stuff out of Gantt class. Refs #9024, #9515, #8982.

[Chris Nelson]

(In [11072]) Refactor scheduling out of PM class. Refs #8982, #9042, #9515.

[Chris Nelson]

(In [11073]) Move Project Management stuff into its own module. Refs #8982, #9042, #9515.

[Chris Nelson]

Plugin now does scheduling based on dependencies.