1 Ikiwiki currently only has one type of dependency between pages
2 (plus wikilinks special cased in on the side). This has resulted in various
3 problems, and it's seemed for a long time to me that ikiwiki needs to get
4 smarter about what types of dependencies are supported.
8 The current single dependency type causes the depending page to be rebuilt
9 whenever a matching dependency is added, removed, or *modified*. But a
10 great many things don't care about the modification case, and often cause
11 unnecessary page rebuilds:
13 * map only cares if the pages are added or removed. Content change does
14 not matter (unless show=title is used).
15 * brokenlinks, orphans, pagecount, ditto (generally)
16 * inline in archive mode cares about page title, author changing, but
17 not content. (Ditto for meta with show=title.)
18 * Causes extra work when solving the [[bugs/transitive_dependencies]]
21 ### two types of dependencies needed for [[tracking_bugs_with_dependencies]]
23 >> it seems that there are two types of dependency, and ikiwiki
24 >> currently only handles one of them. The first type is "Rebuild this
25 >> page when any of these other pages changes" - ikiwiki handles this.
26 >> The second type is "rebuild this page when set of pages referred to by
27 >> this pagespec changes" - ikiwiki doesn't seem to handle this. I
28 >> suspect that named pagespecs would make that second type of dependency
29 >> more important. I'll try to come up with a good example. -- [[Will]]
31 >>> Hrm, I was going to build an example of this with backlinks, but it
32 >>> looks like that is handled as a special case at the moment (line 458 of
33 >>> render.pm). I'll see if I can breapk
34 >>> things another way. Fixing this properly would allow removal of that special case. -- [[Will]]
36 >>>> I can't quite understand the distinction you're trying to draw
37 >>>> between the two types of dependencies. Backlinks are a very special
38 >>>> case though and I'll be suprised if they fit well into pagespecs.
41 >>>>> The issue is that the existential pagespec matching allows you to build things that have similar
42 >>>>> problems to backlinks.
43 >>>>> e.g. the following inline:
45 \[[!inline pages="define(~done, link(done)) and link(~done)" archive=yes]]
47 >>>>> includes any page that links to a page that links to done. Now imagine I add a new link to 'done' on
48 >>>>> some random page somewhere - a page which some other page links to which didn't previously get included - the set of pages accepted by the pagespec, and hence the set of
49 >>>>> pages inlined, will change. But, there is no dependency anywhere on the page that I altered, so
50 >>>>> ikiwiki will not rebuild the page with the inline in it. What is happening is that the page that I altered affects
51 >>>>> the set of pages matched by the pagespec without itself being matched by the pagespec, and hence included in the dependency list.
53 >>>>> To make this work well, I think you need to recognise two types of dependencies for each page (and no
54 >>>>> special cases for particular types of links, eg backlinks). The first type of dependency says, "The content of
55 >>>>> this page depends upon the content of these other pages". The `add_depends()` in the shortcuts
56 >>>>> plugin is of this form: any time the shortcuts page is edited, any page with a shortcut on it
57 >>>>> is rebuilt. The inline plugin also needs to add dependencies of this form to detect when the inlined
58 >>>>> content changes. By contrast, the map plugin does not need a dependency of this form, because it
59 >>>>> doesn't actually care about the content of any pages, just which pages it needs to include (which we'll handle next).
61 >>>>> The second type of dependency says, "The content of this page depends upon the exact set of pages matched
62 >>>>> by this pagespec". The first type of dependency was about the content of some pages, the second type is about
63 >>>>> which pages get matched by a pagespec. This is the type of dependency tracking that the map plugin needs.
64 >>>>> If the set of pages matched by map pagespec changes, then the page with the map on it needs to be rebuilt to show a different list of pages.
65 >>>>> Inline needs this type of dependency as well as the previous type - This type handles a change in which pages
66 >>>>> are inlined, the previous type handles a change in the content of any of those pages. Shortcut does not need this type of
67 >>>>> dependency. Most of the places that use `add_depends()` seem to need this type of dependency rather than the first type.
69 >>>>>> Note that inline and map currently achieve the second type of dependency by
70 >>>>>> explicitly calling `add_depends` for each page the displayed.
71 >>>>>> If any of those pages are removed, the regular pagespec would not
72 >>>>>> match them -- since they're gone. However, the explicit dependency
73 >>>>>> on them does cause them to match. It's an ugly corner I'd like to
74 >>>>>> get rid of. --[[Joey]]
76 >>>>> Implementation Details: The first type of dependency can be handled very similarly to the current
77 >>>>> dependency system. You just need to keep a list of pages that the content depends upon. You could
78 >>>>> keep that list as a pagespec, but if you do this you might want to check that the pagespec doesn't change,
79 >>>>> possibly by adding a dependency of the second type along with the dependency of the first type.
81 >>>>>> An example of the current system not tracking enough data is
82 >>>>>> described in [[bugs/transitive_dependencies]].
85 >>>>> The second type of dependency is a little more tricky. For each page, we'd need a list of pagespecs that
86 >>>>> the page depended on, and for each pagespec you'd want to store the list of pages that currently match it.
87 >>>>> On refresh, you'd need to check each pagespec to see if the set of pages that match it has changed, and if
88 >>>>> that set has changed, then rebuild the dependent page(s). Oh, and for this second type of dependency, I
89 >>>>> don't think you can merge pagespecs. If I wanted to know if either "\*" or "link(done)" changes, then just checking
90 >>>>> to see if the set of pages matched by "\* or link(done)" changes doesn't work.
92 >>>>> The current system works because even though you usually want dependencies of the second type, the set of pages
93 >>>>> referred to by a pagespec can only change if one of those pages itself changes. i.e. A dependency check of the
94 >>>>> first type will catch a dependency change of the second type with current pagespecs.
95 >>>>> This doesn't work with backlinks, and it doesn't work with existential matching. Backlinks are currently special-cased. I don't know
96 >>>>> how to special-case existential matching - I suspect you're better off just getting the dependency tracking right.
98 >>>>> I also tried to come up with other possible solutions: e.g. can we find the dependencies for a pagespec? That
99 >>>>> would be the set of pages where a change on one of those pages could lead to a change in the set of pages matched by the pagespec.
100 >>>>> For old-style pagespecs without backlinks, the dependency set for a pagespec is the same as the set of pages the pagespec matches.
101 >>>>> Unfortunately, with existential matching, the set of pages that each
102 >>>>> pagespec depends upon can quickly become "*", which is not very useful. -- [[Will]]
106 I propose the following. --[[Joey]]
108 * Add a second type of dependency, call it an "presence dependency".
109 * `add_depends` defaults to adding a regular ("full") dependency, as
110 before. (So nothing breaks.)
111 * `add_depends($page, $spec, presence => 0)` adds an presence dependency.
112 * `refresh` only looks at added/removed pages when resolving presence
115 This seems straightforwardly doable. I'd like [[Will]]'s feedback on it, if
116 possible. The type types of dependencies I am proposing are not identical
117 to the two types he talks about above, but I hope are close enough that
120 This doesn't deal with the stuff that only depend on the metadata of a
121 page, as collected in the scan pass, changing. But it does leave a window
122 open for adding such a dependency type later.
126 I implemented the above in a branch.
127 [[!template id=gitbranch branch=origin/dependency-types author="[[joey]]"]]
129 Then I found some problems:
131 * Something simple like pagecount, that seems like it could use a
132 presence dependency, can have a pagespec that uses metadata, like
133 `author()` or `copyright()`.
134 * pagestats, orphans and brokenlinks cannot use presence dependencies
135 because they need to update when links change.
137 Now I'm thinking about having a special dependency look at page
138 metadata, and fire if the metadata changes. And it seems links should
139 either be included in that, or there should be a way to make a dependency
140 that fires when a page's links change. (And what about backlinks?)
142 It's easy to see when a page's links change, since there is `%oldlinks`.
143 To see when metadata is changed is harder, since it's stored in the
144 pagestate by the meta plugin. Also, there are many different types of
145 metadata, that would need to be matched with the pagespecs somehow.
147 Quick alternative: Make add_depends look at the pagespec. Ie, if it
148 is a simple page name, or a glob, we know a presence dependency
149 can be valid. If's more complex, convert the dependency from
152 There is a lot to dislike about this method. Its parsing of the pagespec,
153 as currently implemented, does not let plugins add new types of pagespecs
154 that only care about presence. Its pagespec parsing is also subject to
155 false negatives (though these should be somewhat rare, and no false
156 positives). Still, it does work, and it makes things like simple maps and
157 pagecounts much more efficient.
161 #### Will's first pass feedback.
163 If the API is going to be updated, then it would be good to make it forward compatible.
164 I'd like for the API to be extendible to what is useful for complex pagespecs, even if we
165 that is a little redundant at the moment.
167 My attempt to play with this is in my git repo. [[!template id=gitbranch branch=origin/depends-spec author="[[will]]"]]
168 That branch is a little out of date, but if you just look at the changes in IkiWiki.pm you'll see the concept I was looking at.
169 I added an "add_depends_spec()" function that adds a dependency on the pagespec passed to it. If the set of matched pages
170 changes, then the dependent page is rebuilt. At the moment the implementation uses the same hack used by map and inline -
171 just add all the pages that currently exist as traditional content dependencies.
173 > As I note below, a problem with this approach is that it has to try
174 > matching the pagespec against every page, redundantly with the work done
175 > by the plugin. (But there are ways to avoid that redundant matching.)
178 Getting back to commenting on your proposal:
180 Just talking about the definition of a "presence dependency" for the moment, and ignoring implementation. Is a
181 "presence dependency" supposed to cause an update when a page disappears? I assume so. Is a presence dependency
182 supposed to cause an update when a pages existence hasn't changed, but it no longer matches the pagespec.
183 (e.g. you use `created_before(test_page)` in a pagespec, and there was a page, `new_page`, that was created
184 after `test_page`. `new_page` will not match the spec. Now we'll delete and then re-create `test_page`. Now
185 `new_page` will match the spec, and yet `new_page` itself hasn't changed. Nor has its 'presence' - it was present
186 before and it is present now. Should this cause a re-build of any page that has a 'presence' dependency on the spec?
188 > Yes, a presence dep will trigger when a page is added, or removed.
190 > Your example is valid.. but it's also not handled right by normal,
191 > (content) dependencies, for the same reasons. Still, I think I've
192 > addressed it with the pagespec influence stuff below. --[[Joey]]
194 I think that is another version of the problem you encountered with meta-data.
196 In the longer term I was thinking we'd have to introduce a concept of 'internal pagespec dependencies'. Note that I'm
197 defining 'internal' pagespec dependencies differently to the pagespec dependencies I defined above. Perhaps an example:
198 If you had a pagespec that was `created_before(test_page)`, then you could list all pages created before `test_page`
199 with a `map` directive. The map directive would add a pagespec dependency on `created_before(test_page)`.
200 Internally, there would be a second page-spec parsing function that discovers which pages a given pagespec
201 depends on. As well as the function `match_created_before()`, we'd have to add a new function `depend_created_before()`.
202 This new function would return a list of pages, which when any of them change, the output of `match_created_before()`
203 would change. In this example, it would just return `test_page`.
205 These lists of dependent pages could just be concatenated for every `match_...()` function in a pagespec - you can ignore
206 the boolean formula aspects of the pagespec for this. If a content dependency were added on these pages, then I think
207 the correct rebuilds would occur.
209 In all, this is a surprisingly difficult problem to solve perfectly. Consider the following case:
217 > Doesn't shave self.
221 > [!include pages="!link(ShavesSelf)"]
223 Does ShavedByBob.mdwn include itself?
225 (Yeah - in IkiWiki currently links are *not* included by include, but the idea holds. I had a good example a while back, but I can't think of it right now.)
231 > I have also been thinking about some sort of analysis pass over pagespecs
232 > to determine what metadata, pages, etc they depend on. It is indeed
233 > tricky to do. More thoughts on influence lists a bit below. --[[Joey]]
235 >> The big part of what makes this tricky is that there may be cycles in the
236 >> dependency graph. This can lead to situations where the result is just not
237 >> well defined. This is what I was trying to get at above. -- [[Will]]
239 >>> Hmm, I'm not seeing cycles be a problem, at least with the current
240 >>> pagespec terms. --[[Joey]]
244 ### Link dependencies
246 * `add_depends($page, $spec, links => 1, presence => 1)`
247 adds a links + presence dependency.
248 * Use backlinks change code to detect changes to link dependencies too.
249 * So, brokenlinks can fire whenever any links in any of the
250 pages it's tracking change, or when pages are added or
252 * To determine if a pagespec is valid to be used with a links dependency,
253 use the same set that are valid for presence dependencies. But also
254 allow `backlinks()` to be used in it, since that matches pages
255 that the page links to, which is just what link dependencies are
260 ### the removal problem
262 So far I have not addressed fixing the removal problem (which Will
265 Summary of problem: A has a dependency on a pagespec such as
266 "bugs/* and !link(done)". B currently matches. Then B is updated,
267 in a way that makes A's dependency not match it (ie, it links to done).
268 Now A is not updated, because ikiwiki does not realize that it
269 depended on B before.
271 This was worked around to fix [[bugs/inline_page_not_updated_on_removal]]
272 by inline and map adding explicit dependencies on each page that appears
273 on them. Then a change to B triggers the explicit dep. While this works,
274 it's 1) ugly 2) probably not implemented by all plugins that could
275 be affected by this problem (ie, linkmap) and 3) is most of the reason why
276 we grew the complication of `depends_simple`.
278 One way to fix this is to include with each dependency, a list of pages
279 that currently match it. If the list changes, the dependency is triggered.
281 Should be doable, but may involve more work than
282 currently. Consider that a dependency on `bugs/*` currently
283 is triggered by just checking until *one* page is found to match it.
284 But to store the list, *every* page would have to be tried against it.
285 Unless the list can somehow be intelligently updated, looking at only the
290 Found a further complication in presence dependencies. Map now uses
291 presence dependencies when adding its explicit dependencies on pages. But
292 this defeats the purpose of the explicit dependencies! Because, now,
293 when B is changed to not match a pagespec, the A's presence dep does
296 I didn't think things through when switching it to use presence
297 dependencies there. But, if I change it to use full dependencies, then all
298 the work that was done to allow map to use presence dependencies for its
299 main pagespec is for naught. The map will once again have to update
300 whenever *any* content of the page changes.
302 This points toward the conclusion that explicit dependencies, however they
303 are added, are not the right solution at all. Some other approach, such as
304 maintaining the list of pages that match a dependency, and noticing when it
309 ### pagespec influence lists
311 I'm using this term for the concept of a list of pages whose modification
312 can indirectly influence what pages a pagespec matches.
314 > Trying to make a formal definition of this: (Note, I'm using the term sets rather than lists, but they're roughly equivalent)
316 > * Let the *matching set* for a pagespec be the set of pages that the pagespec matches.
317 > * Let a *complete influence set* for a pagespec be the set of all pages whose alteration might change the matching set of that pagespec.
318 > * Let the *direct influence set* be the intersection of the matching set and the complete influence set.
319 > * Let the *indirect influence set* be the compliment of the direct influence set with respect to the complete influence set.
321 > Is that a fair definition? I don't think it quite matches your examples below unfortunately.
322 > I was unsure if I should insert the word 'existing' in there in a few places. As it stands, these definitions could include sets of pages that don't exist, e.g. "*".
323 > The one I'm least sure of is the definition of the direct influence set. It feels like you want something
324 > like "the traditional set of things we thought about that could cause a pagespec to change", but that definition
325 > is not very formal and I suspect will lead to problems. Something like "The set of pages matched by the globs in the pagespec" might be closer?
328 >> I appreciate the formalism!
330 >> Only existing pages need to be in these sets, because if a page is added
331 >> in the future, the existing dependency code will always test to see
332 >> if it matches. So it will be in the maching set (or not) at that point.
334 >> The problem with your definition of direct influence set seems to be
335 >> that it doesn't allow `link()` and `title()` to have as an indirect
336 >> influence, the page that matches. But I'm quite sure we need those.
341 * The pagespec "created_before(foo)" has an influence list that contains foo.
342 The removal or (re)creation of foo changes what pages match it.
344 * The pagespec "foo" has an empty influence list. This is because a
345 modification/creation/removal of foo directly changes what the pagespec
348 * The pagespec "*" has an empty influence list, for the same reason.
349 Avoiding including every page in the wiki into its influence list is
352 * The pagespec "title(foo)" has an influence list that contains every page
353 that currently matches it. A change to any matching page can change its
354 title, making it not match any more, and so the list is needed due to the
357 * The pagespec "backlink(index)" has an influence list
358 that contains index (because a change to index changes the backlinks).
360 * The pagespec "link(done)" has an influence list that
361 contains every page that it matches. A change to any matching page can
362 remove a link and make it not match any more, and so the list is needed
363 due to the removal problem.
365 >> Why doesn't this include every page? If I change a page that doesn't have a link to
366 >> 'done' to include a link to 'done', then it will now match... or is that considered a
367 >> 'direct match'? -- [[Will]]
369 >>> The regular dependency calculation code will check if every changed
370 >>> page matches every dependency. So it will notice the link was added.
373 #### Low-level Calculation
375 One way to calculate a pagespec's influence would be to
376 expand the SuccessReason and FailReason objects used and returned
377 by `pagespec_match`. Make the objects be created with an
378 influence list included, and when the objects are ANDed or ORed
379 together, combine the influence lists.
381 That would have the benefit of allowing just using the existing `match_*`
382 functions, with minor changes to a few of them to gather influence info.
384 But does it work? Let's try some examples:
386 Consider "bugs/* and link(done) and backlink(index)".
388 Its influence list contains index, and it contains all pages that the whole
389 pagespec matches. It should, ideally, not contain all pages that link
390 to done. There are a lot of such pages, and only a subset influence this
393 When matching this pagespec against a page, the `link` will put the page
394 on the list. The `backlink` will put index on the list, and they will be
395 anded together and combined. If we combine the influences from each
396 successful match, we get the right result.
398 Now consider "bugs/* and link(done) and !backlink(index)".
400 It influence list is the same as the previous one, even though a term has
401 been negated. Because a change to index still influences it, though in a
404 If negation of a SuccessReason preserves the influence list, the right
405 influence list will be calculated.
407 Consider "bugs/* and (link(done) or backlink(index))"
408 and "bugs/* and (backlink(index) or link(done))'
410 Its clear that the influence lists for these are identical. And they
411 contain index, plus all matching pages.
413 When matching the first against page P, the `link` will put P on the list.
414 The OR needs to be a non-short-circuiting type. (In perl, `or`, not `||` --
415 so, `pagespec_translate` will need to be changed to not use `||`.)
416 Given that, the `backlink` will always be evalulated, and will put index
417 onto the influence list. If we combine the influences from each
418 successful match, we get the right result.
420 > This is implemented, seems to work ok. --[[Joey]]
422 > `or` short-circuits too, but the implementation correctly uses `|`,
423 > which I assume is what you meant. --[[smcv]]
425 #### High-level Calculation and Storage
427 Naively calculating the full influence list for a pagespec requires trying
428 to match it against every page in the wiki. I'd like to avoid doing such
429 expensive matching redundantly.
431 It may be possible, for some types of pagespecs, to just try matching a
432 single, arbitrary page against it, and know the full influence list has
433 been obtained. It seems to be that case that if a pagespec has any
434 influences, matching any page will return at least one. So if none are
435 returned, we can skip trying other pages.
437 If the influence list does not include the page that was tried, we know
438 that the pagespec does not things like `link()` and `title()`, that are
439 influenced by the page's own content. So it *might* be safe to not try
440 matching any more pages in this case too. I think it would work for all
441 current pagespec terms. There might be a hypothetical term where this
442 optimisation doesn't work. We could add a special case to ensure it can
443 work: If a term declares it is unfluenced by "", then it means it is
444 always influenced by the matching page.
446 Anyway, this seems worth doing: Add a `pagespec_match_all`, which returns a
447 list of all pages in the whole wiki that match the pagespec, and also adds
448 the pagespec as a dependency, and while it's at it, calculates and stores
451 It could have an optional sort parameter, and limit parameter, to control
452 how many items to return and the sort order. So when inline wants to
453 display the 10 newest, only the influence lists for those ten are added.
455 If `pagespec_match_depends` can be used by all plugins, then great,
456 influences are automatically calculated, no extra work needs to be done.
458 If not, and some plugins still need to use `pagespec_match_list` or
459 `pagespec_match`, and `add_depends`, then I guess that `add_depends` can do
460 a slightly more expensive influence calculation.
462 Bonus: If `add_depends` is doing an influence calculation, then I can remove
463 the nasty hack it currently uses to decide if a given pagespec is safe to use
464 with an existence or links dependency.
466 Where to store the influence list? Well, it appears that we can just add
467 (content) dependencies for each item on the list, to the page's
468 regular list of simple dependencies. So, the data stored ends up looking
469 just like what is stored today by the explicit dependency hacks. Except,
470 it's calculated more smartly, and is added automatically.
472 > I've implemented influence calculation in `add_depends`. As expected,
473 > it means rather a lot more work, and makes some things much slower.
474 > Optimisations next.. --[[Joey]]
478 Note that influences can also have types, same as dependency types.
479 For example, "backlink(foo)" has an influence of foo, of type links.
480 "created_before(foo)" also is influenced by foo, but it's a presence
483 > This is an interesting concept that I hadn't considered. It might
484 > allow significant computational savings, but I suspect will be tricky
485 > to implement. -- [[Will]]
487 >> It was actually really easy to implement it, assuming I picked the right
488 >> dependency types of course. --[[Joey]]
projects / ikiwiki / blob