Whooa!
Templates are powerful.
programming this way is really fun.
under the assumption that the parts are logical,
all conceivable combinations of theses parts are bound to be correct
it passes compilation, but the test still fails, since
I've changed the expected semantics of the iteration,
in the light of the insights I've gained during
re-investigation of the IterExplorer.
What I now actually intend is rather to embed a
HierarchyOrientationIndicator into the iterator,
instead of returning a special "bracket" marker
reference to indicate return from a nested scope.
Only a Record payload constitutes a nested scope.
For all other (primitive) values, we return an empty iterator.
When used within ScopeExplorer, this implementation will just
lead to exposing any simple value once, while delving into
iteration of nested scopes
The only substantial change (besides compilation fixes) is
to confine the iteration to *const access*
This is a good thing; the whole Record/GenNode structure
was designed to represent immutable data, necessitating
a dedicated *Mutator* for any reshaping.
seemingly the operator-> was not yet used in any real scenario.
The whole point with IterAdapter is that it uses an opaque "location type",
which is owned by the controlling container. In many cases this will
actually be just a pointer into the container storage, but we
must not assume it is this way. Thus the only way to obtain a
(language) pointer is to dereference the "position type" and
take the address of the result
Initially I intended just to supply an addapter to use
the monadic IterExplorer for this recursive expansion
of GenNode contents. Investigating this approach was
relevant to highlight the minimum requirements for
such an evaluation mechanics: since our GenNode
is an hierarchical structure without back-links,
we are bound to use a stack at some point. And
since an Iterator is a materialised continuation,
we can not use the processor stack and are forced
to represent this stack in memory.
Yet, on second thought, we do not need the full power
of the IterExplorer monad; especially we do not need
to bind arbitrary functions into the monad, just one
single scope exploring function, implemented as
Variant visitor. Based on these observations, we can
"inline" the monad structure into a double nested
iterator, where the outer capsule carries a stack
of scopes to be explored.
This helper was drafted for the Job / JobPlanning and Scheduler
interface in 2013, but seemingly not yet put into action. While
in the original use case, we have a genuine measuerment for the
tree depth (given by the depth of the processing stack), in other
use cases we want to use to offset embedded within the indicator
itself for keeping track of the depth. Thus I add a second
mark operation, which usess the current offset to set a new
reference level. This has the consequence that the offset
has now to reflect the new reference point immediately
Since C++ is not a real functional programming language and
has unsafe unmanaged pointers, it is not difficult to produce
dangling references within an extended evaluation pipeline
involving transient objects and pass-by-reference.
In the initial implementation, I built in a safeguard copy
into the signature of the Explorer function, to make sure even
a transiently dressed-up input value gets materialised before
proceeding with the source sequence. Unfortunately this safeguard
turns out as a roadblock now; we might as well take the input
by reference and return an "expanded" state by value. We might
even want to do the full "expansion" on referred state, when
we're able to ensure the source values remain in memory
until consumption.
Thus now the full power of decision is placed on the signature
of the explorer function. The expansion strategies of IterExplorer
will no longer attempt to "sanitise" the signature of the passed-in
function to prevent desaster; I've added some warnings into the
documentation to highlight that danger. Basically, if you want
to be clever, then you're bound to read and understand inticacies
of the implementation.
If in doubt, use values and copying. C++ is optimised for that.
allow to pick the bare function signature from any "function like" entity
Note: we're still unsing our own Typelist construct to represent
the function argument types. Since we're now using C++11, this
could be replaced by varargs, and we could get rid of those
various overloads for different numbers of arguments.
Unfortunately this transition is linked to the usage of
argument tuples in our command framework (which could then be
replaced by std::tuple), and this is still a bit of rather
technical work, which I do not want to spend much time on
right now
remembered that some years ago I had to deal with a very similar problem
for planning the frame rendering jobs. It turned out, that the
iterator monad developed for this looks promising for our task at hand
horay!
seems like madness?
well -- found and squashed a bug: equality on RecordRef
implicitly converted to GenNode(RecordRef), which always
generates new (distinct) IDs and so never succeeds. What
we really want is equality test on the references
contrary to the Visitor, accepting a Predicate is const,
and -- of course -- the Predicate invocation returns bool.
This can be used to implement comparison operators or
search functions on Variant based data structures.
There is no generic implementation for these functions, since
they are highly dependent on the payload used within Record<TY>
Here we use Record<GenNode>, which turns the whole setup into an
recursive data type; we especially rely on the fact that each
GenNode has an embedded symbolic ID, and we use this ID to encode
the 'key' for named attributes
while in debugging, it turned out that the short type-prefix
was implemented in a too simplistic way; it fails on stuff
like 'lib::diff::Record<lib::diff::GenNode>'
while I must add, that the whole purpose of these ID functions
is somewhat unclear and needs to reveal itself as we move forward
initially my intention was to use the ID for equality test.
But on a second thought, this seemed like a bad idea, since
it confuses the concepts of equality and identity.
Note: at the moment, I do not know if we even need an equality test,
so it is provided here rather for sake of completeness. And this
means even more that we want an 'equality' implementation that
does what one would naively expect: compare the object identity
*and* compare the contents.
...while on the train back from FrOSCon.
still the same old problem: we need a better hash function
for generating our Entry-IDs. The default hash function from Boost performs
poor on strings with common prefix and trailing number.
We use a hackish workaround, which is sufficient to avoid collisions
among the first 10000 numbers.
not entirely sure about the design, but lets try this approach:
they can be "cloned" and likewise move-assigned, but we do not
allow the regular assignment, because this would enable to use
references like pointers (what we deliberately do not want)
especially setting (changing) attributes turned out to be tricky,
since in case of a GenNode this would mean to re-bind the hash ID;
we can not possibly do that properly without knowing the type of the payload,
and by design this payload type is opaque (erased).
As resort, I changed the semantics of the assign operation:
now it rather builds a new payload element, with a given initialiser.
In case of the strings, this ends up being the same operation,
while in case of GenNode, this is now something entirely different:
we can now build a new GenNode "in place" of the old one, and both
will have the same symbolic ID (attribute key). Incidentally,
our Variant implementation will reject such a re-building operatinon
when this means to change the (opaque) payload type.
in addition, I created a new API function on the Mutator,
allowing to move-in a complete attribute object. Actually this
new function became the working implementation. This way, it is
still possible to emplace a new attribute efficiently (consider
this to be a whole object graph!). But only, if the key (ID)
embedded in the attribute object is already what is the intended
key for this attribute. This way, we elegantly circumvent the
problem of having to re-bind a hash ID without knowing the type seed
initially, the intention was to inject the type as a magic attribute.
But this turned out to make the implementation brittle, asymmetric
and either quite demanding, or inefficient.
The only sane approach would be to introduce a third collection,
the metadata attributes. Then it would be possible to handle these
automatically, but expose them through the iterator.
In the end I decided against it, just the type attribute
allone does not justify that effort. So now the type is an
special magic field and kept apart from any object data.
this solves the problem how to deal with value access
- for the simple default (string) implementation,
we use a 'key = val' syntax and thus have to split strings,
which means we need to return contents by value
- for the actual relevant use case we have GenNode entries,
which may recursively hold further Records. For dealing
with diff messages over this data struture, its a good
idea to allow for const& value access (otherwise we'd
end up copying large subtrees for trivial operaions)
OMG, what was all this about?
OK... this cant possibly work this way.
At least we need to trim after splitting the attributes.
But this is not enough, we want the value, which implies
to make the type flexible (since we cant return a const& to
a substring extracted on-the-fly)
this was an half hearted attempt to satisfy CLang,
but GCC as keen as a razor insists on these inherited
functions not being accessible --
seems like the time is over, when GCC used to be forgiving
and CLang briliantly precise...
So the conclusion of this "round trip" is: whenever GCC
also starts whining about shadowed overloaded virtual functions,
we'll just switch to "-Wno-overloaded-virtual" and be done with
that pointless discussion.
Since C++11, we have the Java style override specifier,
which does a way better job at spotting signature mismatches
Note: not fixing all relevant warnings.
Especially, the "-Woverloaded-virtual" of Clang defeats the whole purpose
of generated generic interfaces. For example, our Variant type is instantiated
with a list of types the variant can hold. Through metaprogramming, this
instantiation generates also an embedded Visitor interface, which has
virtual 'handle(TY)' functions for all the types in question
The client now may implement, or even partially implement this Visitor,
to retrieve specific data out of given Variant instance with unknown conent.
To complain that some other virtual overload is now shaddowed is besides the point,
so we might consider to disable this warning altogether
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=56402
The lambda definition captures the this pointer,
but the ctor of the lamda does not initialise this capture.
In our case, we're lucky, as we don't use the "this" pointer;
otherwise, we'd get a crash a runtime.
Fixed since GCC-4.7.3 --> it's *really* time to upgrade to Debian/Jessie
the object VTable is typically emitted when the compiler
encounters the first non-static non-inline function of
the class or a derived class.
Sometimes this happens within the wrong library and so
the compiler needs a nudge to emit those infrastructure functions.
But in most cases this works out of the box and need no further
magic incanctations, which might have a downside.
Especially because also a non-inline dtor does incur a call overhead,
whereas an inline dtor can be trivially elided.
after sleeping a night over this, it seems obvios
that we do not want to start the build proces "implicitly",
starting from a Record<GenNode>. Rather, we always want
the user to plant a dedicated Mutator object, which then
can remain noncopyable and is passed by reference through
the whole builder chain. Movin innards of *this object*
are moved away a the end of the chain does not pose much risk.
especially I've now decided how to handle const-ness:
We're open to all forms of const-ness, the actual usage decides.
const GenNode will only expose a const& to the data values
still TODO is the object builder notation for diff::Record
forwarding equality to the embedded EntryID
Basically, two GenNodes are equal when they have the same "identity"
Ironically, this is the usual twist with database entities
on a second thought, this "workaround" does not look so bad,
due to the C++11 feature to request the default implementation explicitly.
Maybe we'll never need a generic solution for these cases
I decided to allow for an 'unbound' reference to allow
default construction of elements involving record references.
I am aware of the implications, but I place the focus
on the value nature of GenNode elements; the RecordRef
was introduced only as a means to cary out diff comparisons
and similar computations.
basically this is the well known problem #587
Just it became more pressing with the Upgrade to Jessie and Boost 1.55
So I've pulled off the well known "Knuth trick" to spread the
input data more evenly within the hash domain.
And voilà: now we're able to use 100000 number suffixes without collision
- move the santitise operation up into EntryID's ctor
- turn the recast() operation into a real in-place cast
these changes should be transparent to the existing usages
of EntryID (within the asset framework), but allow for use
as attribute name holder in GenNode, since we're now able
to feed existing name/ID values directly into the ctor
of BareEntryID, without any spurious santitise operation.
this was introduced into namespace mobject and spread from there.
Since the habit is to use more specific typedefs like PClip,
it is preferrable to spell out the full namespace
using the struct-scheme.hpp and the requirements for
EntryID as a guideline. The goal is to move EntryID
over into the support lib, which means we need to get rid
of all direct proc::asset dependencies. Thus, these generic
ID functions shall form a baseline implementation, while
asset::Struct may provide the previously used implementation
through specialisation -- so the behaviour of EntryID will
not change for the structural assets, but we'll get a more
sane and readable default implementation for all other types.
before engaging into the implementation of lib::Record,
I prefer to conduct a round of planning, to get a clearer
view about the requirements we'll meet when extending
our existing list diff to tree structures
Initially, I considered to build an index table like
collection of ordered attributes. But since our actual
use case is Record<GenNode>, this was ruled out in favour
of just a vector<GenNode>, where the keys are embedded
right within the nameID-Field of GenNode.
A decisive factor was the observation, that this design
is basically forced to encode the attribute keys somehow
into the attribute values, because otherwise the whole
collection like initialisation and iteration would break
down. Thus, a fully generic implementation is not possible,
and a pseudo generic implementation just for the purpose of
writing unit tests would be overkill.
Basically this decision means that Record requires an
explicit specialisation to implement the attribute-key
binding for each value type to use.
Ouch!
Why does C++ lack the most basic everyday stuff?
It needn't be performant. It needn't support some fancy
higher order container. Just join the f***ing strings.
use Bosst?? -- OMG!! pulls in half the metra programming library
and tries to work on any concievable range like object. Just
somehow our Lumiera Forward Iterators aren't "range-like" enough
for boost's taste.
Thus let's code up that fucking for-loop ourselves, once and forever.
This is kind of the logic consequence, since we consider our
functional iterator concept still superior and will continue
to rely on it.
For some time now, I've considered to build a generic bridge
function, to use enable_if and metaprogramming to figure out
if some type is a "Lumiera Forward Iterator" automatically.
But since our concept is to some degree a contract regarding
semantics, which never can be captured by any kind of introspection,
such a bridge implementation would be rather heuristic and
bears the danger to trigger on types actually not intended
as iterator at all. So I consider such a solution as dangerous
and we'll settle with just supplying the necessary bridge
functions as free functions injected for ADL on a case by case base
this was spotted by a new GCC warning -Wunused-function
and I must admit, GCC is right here: an externally not visible
function in an anonymous namespace is not what I'd expect to be
picked up by ADL. It is rather weird that the metaprogramming
trait worked at all.
Note that the function is intentionally declared only, never defined.
We want a linker error in case boost::hash ever attempts to
use this 'deliberately ill-definded' catch-all.
I'd never imagine that this superficial draft will sit there
for 5+ years without me getting any chance to continue with that topic.
this is so saddening, so I turned off the warning :-/
In Lumiera, "Tracks" are not what you'd expect from
conventional video editing software. They are a mere
grouping devide, and are also used to implement the
"media bins" and tool palettes.
But having "folders" on the timeline would be likewise
confusing, as would be to have a "branch" or "tree".
To get out of that dilemma, we chose an understandable
but deliberately somewhat strange name: "Fork"
It was common understanding on the Mailinglist that we
should handle this renaming in a tuned-down and discrete
way: The UI will continue to show "Tracks" for a familiar
sight and "Bins" in the Asset section. But Lumiera developers
will be nudged to accomodate by renaming the entity in
source code accordingly
Cockoo hashing is a thrilling algorithm.
We investigated it during the time or our first draft
towards a confirugation system in 2008. This usage turned
up some problems -- not sure if based on the implementation
or the algorithm itself; at that time, we just switched
to the probabilistic splay tree. The whole configuration
system effort stalled afterwards; so the cuckoo implementation
remained in tree as a zombie.
This switches the Lumiera UI from GTK-2 to GTK-3
Unfortunately, this move breaks two crucial features, which have been
disabled for now: the display of video and our custom timeline widget.
Since both of these require some reworking, which in fact has already
started, we prefer to do the library and framework switch right away.
over time, a specific Lumiera code writing style has emerged.
The GUI, as it stood, used somewhat different conventions,
which now have been aligned to the common standard.
Basically we use GNU style, with some adjustments for OO-programming,
we prefer CamelCase, and write TypeNames uppercase, variableNames lowercase
it is a widely accepted rule to shape names with the usage site in mind.
Especially this means, that we use the singular form for all kinds
of collections and assortments.
Thus, the namespace should be called "widget" not "widgets",
because at usage site this becomes gui::widget::TimelineWidget
Likewise for "dialogs" and "pannels"
a long standing TODO to document the actual start-up sequence, which
is implemented this way since a long time now. There was an unwritten
section in the "Linking and Application Structure", which seems the
apropriate place for this kind of intricate techincal details.
Last week, Benny Lyons was here on visit in munich and he was pondering
the idea of an experimental secondary build system, as a way to learn
more about the source structure of Lumiera. This reminded me to fill
some missing parts of the documentation. Possibly this is also the
right moment to land the GTK-3 transition?
The actual trick to make it work is to use decltype on the function operator
http://stackoverflow.com/questions/7943525/is-it-possible-to-figure-out-the-parameter-type-and-return-type-of-a-lambda/7943765#7943765
In addition, we now pick up the functor by template type and
store it under that very type. For one, this cuts the size
of the generated class by a factor of two. And it gives the
compiler the ability to inline a closure as much as is possible,
especially when the created Binder / Mutator lives in the same
reference frame the closure taps into.
to carry out that rather obvious step, I was bound to consider
all the implications of choosing a given layout and handling pattern
for our external structure representation.
Finally, I settled upon the following decisions
- the value space represented within the DataCap is flat, not further structured
- the distinction between "attribute" and "nested object" is merely conceptual
and will be enforced solely by the diff detection / representation protocol
- basically, a nested subtree may appear as an attribute; the difference
between attributes and children lies solely in the way of access and referral:
by-name vs. positional
- it is pointless to save space for the representation of the discriminator ID
- but we can omit any further explicit type tag, because
- we do *not* support programming by switch-on-type, and thus
- we do *not* support full introspection, only a passive type-safety check
- this is *not* a limitation, since we acknowledge that GenNode is a *Monad*
- and the partial function needed within any flatMap implementation
maps naturally onto our Variant-Visitor; thus
- the DataCap can basically just *be* a Variant
- and GenNode has just to supply the neccessary shaffolding
to turn that into a full fledged Monad implementation, including
direct construction by wrapping a value and flatMap with tree walk
All relevant uses will rely on the more strict access policy
implemented with the new util::AccessCasted. Along the same line
of thinking, I've removed the "second try" convenience conversion
from the typed get-Function of OpaqueHolder. Such an unbounded
"convert it somehow" approach is almost never a good idea. Either,
one knows by design the precise type to expect, or alternatively
should rely on the base interface solely.
...with the sole exception of the usage in WrapperPointer,
which in itself looks obsolete to me; we should better re-think
the way we handle "wrapped" objects for the BuilderTools, once
we actually start implementing the Builder
Ticket #450
Note: the new Variant implementation is a re-write from scratch
and does not rely on util::AccessCasted any more. Anyway, both
are now thoroughly covered by unit test
NOTE: this was a one-time verification. Unfortunately there is no way
to verify a failing compilation automatically from a unit-test.
Thus we need to comment out these invalid cases, leaving them
here just for later referral. Need to check those manually
for new compilers to be sure!
this overload will be picked only if none of the more specific
overloads is applicable. Instantiating this overload will then
trigger a static assertion failure. This way we sort out
impossible or dangerous combinations at compile time already.
I found no simple way to include the actual type parameters in
the generated error message (string concatenation at compiletime)
The throw-statement is only there to prevent a warning due
to missing return statement.
...since I consider that a comparatively safe convenience feature.
Of course we *do perform* a NULL check and throw an exception.
So now the actual casting or conversion functions are designed
to work always on the same level of references or pointers,
which means we can just use the standard conversions of the
language. This has the nice effect of ruling out dangerous
combinations (like taking a L-ref from a R-ref) automatically
TODO: might break some unit-tests...
Explanation: our wrapper around boost::format has special
built-in support for custom operator string(). Any type,
which is neiter standard, or printable through such a
custom string conversion, is represented as a type-string.
For this fallback case, we now use our recently added
demangling call (which actually relies on a rather obscure
but standard compiler API)
still passes compilation, but not actually tested.
The visitor-style accees needs to be implemented, and the
whole virtual copy support mechanism extracted into a separate
header and covered by unit test
now the solution with the copy policy class is in place,
I prefer to return to the more verbose yet clearer notion
of distinct constructors for each case on the outer and
the inner capsule likewise.
The idea with the separate builder class would be significant
only if this class would also provide the copy support. This
turns out to be difficult, due to the access restrictions
and the necessary passing of type parameters.
turns out to be quite a tough challenge....
since obviously we want to support usage of types with
partially disabled copy/assignment operations within Variant.
As long as the corresponding operations on the container aren't
invoked, we expect those types to be usable just fine.
The problem arises at the interaction with type erasure;
to support corret copy / assignement in such a situation, we need
virtual copy / assignment operators. And, since these are to be installed
into a VTable, the templated functions will be instantiated allways,
which might cause invocation of inhibited copy / assignement functions
and thus compilation failure, in spite of never actually invoking such
an illegal operation.
The drafted solution is to mix in a specifically configured copy support policy,
which at least raises a runtime error, instead of invoking the incriminating operation(s)
finally got all those copy / assgnment flavours straight.
Still unsolved: unable to instantiate the Variant template
for a type with private assignment operator (like e.g. Time )
The problem is our virtual assignement operator, which forces
instantiation of the implementation (for the VTable), even if
the actual assignment is never invoked.
this was an immature first desgin attempt; we need a lightweight
Variant (typesafe union) implementation, so now is the time for
a second attempt. The existing Variant is used only once, and this
usage as such is in a questionable context, likely to be reworked
when we actually start coding up the builder. So I'll just move
it away and mark it @deprecated for the time being.
After some reconsideration, I decide to stick to the approach with the closures,
but to use a metaprotramming technique to build an inheritance chain.
While I can not decide on the real world impact of storing all those closures,
in theory this approach should enable the compiler to remove all of the
storage overhead. Since, when storing the result into an auto variable
right within scope (as demonstrated in the test), the compiler
sees the concrete type and might be able to boil down the actual
generated virtual function implementations, thereby inlining the
given closures.
Whereas, on the other hand, if we'd go the obvious conventional route
and place the closures into a Map allocated on the stack, I wouldn't
expect the compiler to do data flow analysis to prove this allocation
is not necessary and inline it away.
NOTE: there is now guarantee this inlining trick will ever work.
And, moreover, we don't know anything regarding the runtime effect.
The whole picture is way more involved as it might seem at first sight.
Even if we go the completely conventional route and require every
participating object to supply an implementation of some kind of
"Serializable" interface, we'll end up with a (hand written!)
implementation class for each participating setup, which takes
up space in the code segment of the executable. While the closure
based approach chosen here, consumes data segment (or heap) space
per instance for the functors (or function pointers) representing
the closures, plus code segment space for the closures, but the
latter with a way higher potential for inlining, since the closure
code and the generated virtual functions are necessarily emitted
within the same compilation unit and within a local (inline, not
publickly exposed) scope.
so yes, it is complicated, and inevitably involves three layers
of indirection. The alternative seems to bind the GUI direcly to
the Session interface -- is there a middle gound?
For the messages from GUI to Proc, we have our commands, based
on PlacementRef entities. But for feeding model updates to the
GUI, whatever I consider, I end up either with diff messages or
an synchronised access to Session attributes, which ties the
responsiveness of the GUI to the Builder operation.
- we use a GenNode element
- this holds a polymorphic value known as DataCap
- besides simple attribute values, this may hold collections of GenNode sub elements
- a special kind of GenNode collection, the Record, is used to represent objects
The purpose of this setup is to enable an external model representation
which is only loosely coupled to the interndal data representation
through the exchange of (tree)diff messages
previously this operation was named 'attach', which an be confused
with attching an object to this location. Indeed, the session interface
even offers such an attach function. By renaming the focus moving
operation into QueryFocus::shift(Scope), this ambiguity is resolved
This is the first step towards a generic backbone to connect
any GUI elements to the session within Proc-Layer.
It is based on a spefic understanding of Model-View-Controller,
which turns the Model-Controller interactions into messages.
sans the implementation of the index lookup table(s)
The algorithm is KISS, a variant of insertion sort, i.e.
worst time quadratic, but known to perform well on small data sets.
The mere generation of the diff description is O(n log n), since
we do not verify that we can "find" out of order elements. We leave
this to the consumer of the diff, which at this point has to scan
into the rest of the data sequence (leading to quadratic complexity)
finally....
The problem is that the C++ "dependent types" defeat the typical
DSL usage, where you define some helper function in a generic
language setup class and mix this language in as superclass.
This is, C++ requires us to refer explicitly to any dependent type,
since, due to possible template specialisations, the parser
can't know if a given symbol is a inherited type or a field.
As a solution, we place the token constructor functors into a
static struct "token", which allows to write e.g. token.insert(xyz)
As decided in beb57cde
this changeset switches our basic list diff language to work
in the style of an insertion sort. Rather than 'pushing back'
out-of-order elements, we scan and bring forward missing elements.
Later, when passing the original location of the elements
fetched this way, a 'skip' verb will help to clean up
possible leftowers, so implementation is possible
(and indeed acomplished) without shifting any other elements.
and this adds a twist: conceptually, we identify the token
with the abstract handler function it represents. But C++
does not allow us to compare member pointers to virtual functions,
for good reason: even two pointers with the "same offset" into
the VTable might end up referring to different implementations,
when bound to instances of different subclasses. This is what
polymorphism is all about.
At this point it seems reasonably, albeit a bit uggly, to use the
diagnostic ID as placeholder instead, and just compare these IDs
instead. We assume that in practice tokens will be defined through
the provided helper macro, which ensures unique identifiers.
basically just a function to pick up the container and element type automatically.
The actual implementation is delegated to the exisiting lib::iter_stl::IterSnapshot
Heureka! found out that the C++ standard library exposes a
cross vendor C++ ABI, which amongst others allows to show
object code names and type-IDs in the language-level, human
readable unmangeld form.
Of course, actual application code should not rely on such a
internal representation, yet it is of tremendous help when
writing and debugging unit tests.
Signed-off-by: Ichthyostega <prg@ichthyostega.de>
Actually I arried at the conclusion, that the *receiving* of
a diff representation is actually a typical double-dispatch situation.
This leads to the attempt to come up with a specialised visitor
as standard pattern to handle and apply a diff. Obviously,
we do not want the classical GoF-Visitor, but (yes, we had
that discussion allready) -- well in terms of runtime cost,
we have to deal with at least two indirections anyway;
so now I'm exploring the idea to implement one of these
indirections through a functor object, which at the same time
acts as "Tag" in the diff representation language (instead
of using an enum as tag)
Uniform sequence at start of source files
- copyright claim
- license
- file comment
- header guard
- lumiera includes
- library / system includes
Lumiera uses Brittish spelling. Add an according note to the styleguide.
initial considerations; there is a concurrency problem, since
all of session handling within Proc is deliberately not threadsafe.
Thus the decision is to make this the gui::model::SessionFacade's responsibility
The actual problem is not resolved; the pluginloader
should detect the duplicate and not add the handle
to the database initially. Or it should add it
as "duplicate" or "alternate implementation"
Which probably means we need to coder some additional
corner cases. But certainly not now, we have other
more important stuff to do first... we've already
lost the battle against Duke Nukem Forever :-P
Mark parts of the timeline state handling which will certainly
not be retained: any part where the GUI widgets "hold" some kind
of model. GUI widgets shall be *mapped upon* a model representation
and *wired* with callbacks.
Especially I am suspicious when GUI presentation code "reaches into"
any kind of model data structure to find out something. It should
be the other way round (dont call us, we call you)
actually we should make our timeline a real custom widget,
and do it according to the letter. I.e. really implement
all those callbacks which are recommended, but no other
callbacks.
This has the additional benefit of being able to retrieve
the drawing style in the official way, and define our own
CSS classes, which can be styled by the user in a systematic way.
This is not really a solution, but kind of narrowes down the problem.
Our GUI uses an obsolete C-ish approach at releasing resources at
several points. This is probably a left-over from earlier days.
Especially since we started out with libGDL without C++ wrappers.
And at that time, we didn't use smart-pointers, as we should do,
but we tried to do things manually, which is an approach which never
works in an event driven and condition based environment. Goto fail.
Here I just commented out the manual clean-up code from several dtors.
The real solution would be not to allocate these resources through
the raw C calls at first place, but rather use the mm-wrappers
and leave it to them to unwind at the right moment.
TODO:
- scan the GUI code for *every* instance where we still muck around with gobjects
and either replace that by a mm-wrapper, or wrap it in a smart handle.
- make sure that *all* dtors are either empty, or really airtight and EX_FREE
doh...
this happens when you draft some quite intricate logic and then
get sidelined with other tasks for several years. Mind me, I didn't
even recall that I had treated this whole issue and created
a clean-up thread.
A full fledged implementation would have a real lifecycle and
thus detect the re-entrance; but since none of the components
to be managed by the OutputDirector is even remotely planned
or even coded, the functions were just drafted as stubs.
Which caused us happily to create yet another clean-up thread
whenever the subsystem-runner signalled "please shut down".
Our GUI shutdown logic looks rather confused. Why the hell do
some widgets "unregister" themselves in a dtor. This should never
be necessary. Maybe it's a leftover from C-style programming
and obsolete now, after the switch to GDLmm
the problem seems to be the interplay with the installed user
theme; what looks well with one theme is messed up with another
one, quite insidious.
Attempt to settle down on a set of default definitions
for borders and element background colors, which make the
Lumiera Gui "dark" but still respect the user's theme
for geometry and widget style
verified: basically works
todo: better handling of parse errors.
Currently this is treated as an unexpected exception and just
terminates the whole application, without any suitable diagnostics.
This makes working on the stylesheet somewhat brittle. GTK-3 actually
offers a signal to be invoked in case of CSS parsing errors
(see #953)
Comment out the active part of the GdkDisplayer implementation,
but retain the class, to make compilation pass.
With the Switch to GTK-3, only Cairo drawing is supported.
We need a new solution for video display...
Lots off commented out blocks of code
but most issues are related to simple function name changes,
set/get_flags calls, anything that has to do with a Gtk::Style...
Plan of attack from here is to go one-by one of each commented-out or code and update to gtk3 specs.
trying to track down where these messages
GLib-GObject-CRITICAL **: g_object_unref: assertion G_IS_OBJECT (object) failed
are coming from. These appear when iconifying panels.
as always, it turned out that the alledged "compiler bug"
rather was my own sloppyness: I forgot properly to undo a change
I made while fighting with compilation problems: the wrapper in
the factory didn't use std::forward, resulting in a plain flat
slicing copy. This, rightfully, triggered the assertion in the
session query resolver (since a sliced Goal can not be dynamic cast
to a specific Query subclass).
TODO: the toolfactory needs a redesign anyway,
this was just placeholder code added in a very early
state of the Lumiera project. We have way better memory
managing facilities at hand now
the use of a custom finisihing functor, which is applied
to any generated product. This can be used for registration,
memory management or similar framework aspects
Implement the first simple usage scenario for the
unified MultiFact template, using variadic templates.
NOTE:
- the obvious solution based on std::forward
triggers strange behaviour in GCC-4.7
- the inline lambda in the test case traps the
CLang-3.0 parster with a segfault. Horay!
...but the whole design looks still overengineered. See #388
- should get rid of the explicit specialisation
- always use a function signature and thus have arguments?
- why inheriting from the wrapper?
still puzzled why this instantiation of MultiFact fails to compile with GCC 4.8
so I'm bound to understand why the types involved
need indeed to be are structured the way they are right now.
previous versions used to resolve this ambiguity in favour of a ctor call,
but now the compiler treats such constructs as function definition;
this is reasonable, since C++11 introduced the notion of a "generalised
initialisation", which is always written as a (possibly empty) list
in braces.
In these specific cases here, we just omit the empty parens
Since we have now a generic bridge to supply std::hash functions based on
an exisiting boost::hash function, we removed the explicit std::hash
specialisation for "Placement-ID".
This caused the PlacementIndex implementation to fail at compilation
with a quite obscure error, which in fact seems to be caused by the
absence of any specific specialisation. The symptom is that a iterator
range could not be assigned to the predefined iterator type of
std::unordered_multimap, due to a mismatch in the embedded traits type
__umap_traits<__cache_default<_Key, _Hash>::value>>
While I didn't track down that problem entirely, to verify my hypothesis,
the problem can be avoided by using the default -- which is now to
pick up an existing boost::hash function for this type and use this
to generate the std::hash function.
right now we have to defeat an unfortunate static assertion in
the standard library, which is expected to go away in the future.
We use a hack to hijack the problematic definition with the preprocessor,
which requires our header to be first.
NOTE: this header contains a potentially dangerous, temporary workaround
to defeat the static assertion in the default implementation of std::hash,
as shipped with GCC 4.7.x
This assertion turns out to be detrimental all kinds of metaprogramming
based solutions, since it defeats SFINAE. It is expected to be removed
in GCC 4.8
the rules are:
- our own headers go before any library headers
- all headers need to be spelled relative to include root
- ensure that gtk is always included via gui/gtk-base.hpp
c++11 uses another hashtable implementation.
This uncovered some poorly written tests, which relied on
objects being returned in a specific order. As far as poissible,
we're using generic query functions now to get our test objects.
But these tests still rely on a specifically crafted test index content,
which as such is acceptable IMHO. The only remaining problem is
that we check the order of generated output in some tests, and this
order is still implementation dependent.
a real fix would be to rewrite the test to collect the retrieved
values and do a structural verification of the results. This
would mean to write a lot of code for such a marginal topic,
which was implemented just for sake of completenes anyway.
Hopefully my lack of "motivation" doesn't backfire eventually ;-)
Conversion means automatic conversion. In our case,
what we need ist the ability to *construct* a bool from
our (function) object -- while functors aren't automatically
convertible to bool. Thus we use one of the new predicates
from <type_traits>
...uncovered by switching to c++11
When invoking an individual test, we used to erase
the 0-th cmdline argument, which happens to be allways
the name of the test being invoked. Yet none of our
tests actually complied to that contract. Rather,
all tests taking arguments access them by 1-based
argument index. Previously, the argument values just
happened to be still in memory at the original location
after erasing the 0st element.
"Fixed" that by changing the contract. Now, the 0th argument
remains in place, but when there are no additional arguments,
the whole cmdline is cleared.
This is messy, but the test runer needs to be rewritten
entirely, the whole API is clumsy and dangerous. Ticket #289
* use a development snapshot of lib SigC including the recent C++11 adaptations
* never include whole namespaces. Here we got a clash between std::bind and sigc::bind
* use lambdas
* to make the binding code more readable
* to take the nested invocations apart, which resolves the return type ambiguity
In the November developer meeting, Christian and I agreed that
it's best to remove that offending LUID specifications altogether.
Those embedded LUIDs where one of the issues blocking the transition to C++11
This is a partial and preliminary fix; we had an occasional
numeric overflow on 32bit platforms in some tests.
The complete fix will be to introduce a typedef and then
rework the relevant APIs (which are preliminary anyway,
thus no urge right now)
our front-end for boost::format, the class lib::_Fmt
was lacking an reliable specialisation for long and ulong.
This is due to the notorious problem of these types being
of platform dependant size. As a fix, we're speclialising
explicitly for int16_t, int32_t and int64_t and avoid the
common names 'short', 'int' and 'long' alltogether.
And especially for non-64bit-platform (NONPORTABLE)
we add an explicit specialisation for long
The recommendation is to use the link flag --no-undefined
and to fed *all* dependencies to the respective link step.
This changeset enables this strict linking of dependencies.
It turned out that our dependencies were already sane
(with the sole exception of a direct dependency to X-Lib
in the XV viewer widget)
- upgrade the configuration to a current version
- provide a frontpage with cross-links to other documentation
- define a set of modules; relevant classes and files can be
added to these, to create a exploration path for new readers
- fix a lot of errors in documentation comments
- use a custom configuration for the documentation pages
- tweak the navigation, the sections and further arrangements
to make them stand out more prominently, some entity comments
where started with a line of starts. Unfortunately, doxygen
(and javadoc) only recogise comments which are started exactly
with /**
This caused quite some comments to be ignored by doxygen.
Credits to Hendrik Boom for spotting this problem!
A workaround is to end the line of stars with *//**
When a ctor throws, the dtors of sub-objects have already been
invoked. The object itself never existed, strictly speaking,
and thus the dtor must not be invoked. Usually the runtime system
handles matters automatically this way, but since we're doing
here placement new into an array, we're responsible ourselves
This error was uncovered by compiling with Clang.
GCC automatically neutralised this erroneous dtor invocation.
This removes the central clean-up registry;
Instead, now the InstanceHolder manages the lifecycle of
the service instances placed into static memory; the net effect
is that DependencyFactory and instances are created and destroyed
together, locally for each usage scope
We don't need this ability and it pushes us into using a
central registry. This solution turned out to be problematic
when loading dynamic libraries (plug-ins).
this check may look weird, but in fact a similar check in the
old version of the singleton factory helped us spot a problem
with Clang, most likely but of the compiler or runtime system
Clang doesn't allow to declare a private nested class as friend.
This is unfortunate, but likely correct to the letter of the standard.
As a workaround, now we're creating the instances within a static
function of DependencyFactory -- in the end this improves readability
A second issue fixed with this changeset is the scope of the
marker function. Clang is right, this isn't ADL, thus an inline
friend definition is simply not visible outside the class.
lib::Depend<TY> works as drop-in replacement for lib::Singleton<TY>
This changeset removes the convoluted special cases like
SingletonSub and MockInjector.
explanation: we use pthread_once to define a mutex type descriptor,
used to define some of our mutexes as recursive mutexes. Now,
pthread_once relies on a counter stored in a given location;
we used a non-exported global var for this counter.
Unfortunately this ties the mutex initialisation to the static
initialisation of the compilation unit holding this counter variable.
Theoretically it would be possible (we never observed such an incident)
that, during static initialisation, a singleton was brought up,
which requires a class-scoped lock, implemented as recursive mutex.
And it would be possible for this singleton locking to happen prior
to initialisation of the mentioned counter variable.
As a fix, I've moved the counter varialbe into a function scoped
static variable, since that is guaranteed by the C++ runtime system
to be initialised at first usage of the function, irrespective of the
initialisation order of the enclosing compilation units
clang-3.2 requires a clarification here (while previous versions
of clang and GCC automatically resolved the ambiguity by assuming
use of a nested, dependent template).
Clang is more insistent when it comes to enforcing 'protected' visibility.
Since in this case the basic design can be considered sane and optimal, the
only (and obvious) solution is to nest the PIMPL into a default base class
for implementation; this mirrors the structure of the interface.
Compilation with Clang 3.0 (which is available in Debian/stable) fails,
mostly due to some scoping and naming inconsistencies which weren't detected
by GCC. At some instances, Clang seems to have problems to figure out a
perfectly valid type definition; these can be resolved by more explicit
typing (which is preferrable anyway)
by providing a custom copy function one can adjust otherwise non-copyable
elements. This should be used cautionary because dereferencing elements may
poison the cache and thus have some considerable performance impact
(profile this)
using our util::_Fmt front-end helps to reduce the code size,
since all usages rely on a single inclusion of boost::format
including boost::format via header can cause quite some code bloat
NOTE: partial solution, still some further includes to reorganise
this draft fills in the structure how to get from an invocation
of the engine service to the starting of actual CalcStream instances.
Basically the EngineService implementation is repsonsile to
instruct the Segmentation to provide a suitable Dispatcher.
This is necessary since the implementation of the job functions
calls through the VTable of the interface JobClosure. Thus this
interface (and the VTable definition) needs to reside within
some compilation unit linked together with the basic job class.
TODO: move class Job entirely into the Backend
now builds for me on Debian-7 Wheezy 64bit
unqualified member functions in dependent base classes not found anymore.
Need to qualify either the class or the instance.
...for the very specific situation when we want
to explore an existing data structure, and the
exploration assumes value semantics.
The workaround then is to use pointers as values.
...attempt to build it based on the monadic iterator primitives.
Only problem is: need to find out relation between nodes
after the fact. In the real usage situation, this
is not a problem, since we have a state object
there, which can track the relation as it is established
basically I've changed my mind to prefer an
infinite JobPlanningSequence, which is just
evaluated partially. This removes the need to
embody the logic of planning chunk generation,
which really is a different concern.
decision: the base for any deadline calculations
is the expected real time corresponding to the grid origin.
This value is contained in the Timings record.
this clarifies the relation of TimeAnchor and Timings,
the latter act as a general spec and abstracted grid,
while the latter actually performs the conversion and
deadline checking
Rationale: this is the *support* Library.
The real "Lumiera-Library" does not exist yet.
liblumiera.so will be the *interface* every external
module / plug-in uses to get Lumiera functionality.
Especially the work on Library dependency clean-up
made outright clear, that this interface library
needs to be a separate piece of software, which is
carefully crafted, and more-or-less depends on the
whole application.
the buildsystem will now pick up and link
all test cases according to the layer, e.g.
backend tests will automatically be linked
against the backend + library solely.
tests used to be defined ad hoc and test definitions
are scattered confusingly over various directories.
Now built some simple rules into the buildsystem
to allow organising the tests into layers and
linking them accordingly.
Note: this switches to building shared objects
for the test classes too, which effectively speeds up
both re-building and re-running of test cases
Our libraries constitute a clear dependency hierarchy,
we do not want circular dependencies. Declaring these
dependencies while creating the shared libraries would
allow strict checking by the linker; but unfortunately
this also creates transitive depdendencies stored as
DT_NEEDED tags.
While basically this would be just fine, the resolution of $ORIGIN
on gets confused in case of transitively defined library dependencies
over multiple hops, especially in case when actually no symbol of this
transitive dependency is used. Since these newer systems set the
--as-needed switch for linking by default, these unnecessary
DT_NEEDED entries will be purged from the executable, but of course
not from the shared library causing the transitive dependencies.
As a consequence, when loading the executable, the $ORIGIN resolution
mechanism doesn't act on the dependencies recorded in the library,
causing the shared loader to abort with an "unresolved dependency"
So the resolution for these problems is not to use transitive
dependencies on libraries intended to be found via $ORIGIN
This is kind of a workaround to avoid having to maintain two variants.
Explanation: between Boost 1.42 and 1.52 there was the transition to a
reworked version of the filesystem library, itroducing some breaking changes
The new version distinguishes much clearer between the native and the
generic representation of paths -- which becomes relevant when porting
to non-POXIX operating systems.
Actually the intention was to use the generic path representation in all
configuration; currently this distinction is moot, since we're caring
only for POSIX systems.
So the workaround is to use the fsys::path::string() function, which
is available in both versions, but changed meaning to yield the native
string. Later, when able to deprecate older Boost versions, we should
switch to generic_string()
Note: an alternative solution was found by Mike Fisher in 3b39f35
using the compiletime define BOOST_FILESYSTEM_VERSION=2
See also ticket #896
now this library doesn't refer to any symbols from
Proc-Layer anymore. Resolving these problems
highlighted IMHO a serious shortcomming of our
interface system, which hinders the building
of abstractions at interface level
there is now a mechanism to allow sprcialised queries
to generate this syntactic representation only on demand
The actual concrete representation e.g. for scope queries
still remains TODO, but this won't really change
until we target the integration of a real resoloution engine
while refactoring, I thought it might be a good idea
only to use Query objects. But in this special case,
most often you'd just want to pass in a simple query
with a literal query string. So this convenience shortcut
indeed makes sense.
...to extract the syntetic ordering from
DefsRegistry and make that a responsibility
of the (internal) syntactic representation
of the query.
doesn't pass the compiler yet
effectively this joins the two existing lines
of "Query" classes into one systematic representation
Next step would be to move all mutation operations
over to the Query::Builder
time handling is part of the library, while this
convenience shortcut relies on the Advice system,
which resides in the application lib.
To allow this kind of symbolic acces to a grid
entity defined "elesewhere", client code needs
to be linked against liblumieracore.so
especially this allows to use the Advice system
or the query resolvers from within library facilities
to refer to other implementation level services by name
the rules-based configuration and query system
will be located within the core application,
while the concrete implementation facilities
are expected to reside within the session or
maybe also the GUI.
This is kind of a 'rochade' refactoring to resolve
circular library dependencies and confine the parts
dependant on the session and MObjects to the Proc-Layer
And while we're in the middle of chainsaw surgery,
we'll concentrate further query-based facilities
alongside the config-rules within the App core.
This template was a leftover from the early days
of Lumiera development and doesn't provide any
substantial value as an abstraction.
For the more intricate cases, we're using the
lib::MultiFact template, which allows to install
several "fabrication" functions at runtime
the solution is to introduce a superinterface
and let Dispatcher augment that with the specific parts.
This way, the Job planning only has to rely on the
rather generic stuff (TimeAnchor, FrameCoord)
NOTE: this commit makes the whole JobPlanning machinery
compilable for the first time!
..the Idea is to rely on some kind of service,
to break the cyclic dependency with the Dispatcher.
But I seem unable to find a natural location or
concept to house that service.
basically we had two lines of doubly nested capsules, due to
using the IterAdaptor template. Actually, the evaluation stack
within JobTicket can be considered an implementation detail and
thus doesn't require an iterator interface; the intention is to
use this through JobPlanning solely.
Thus this reworking removes the special iterator within JobTicket,
but retains the idea of exposing the "current" JobTicket through
a smart pointer or operator->()
work done during the FrOSCon travel
especially the exploration stack is pushed down
first successful definition of all the JobPlanning classes
just the framework of classes necessary to pass the compiler;
all implementation is still stubbed
brainstorming how to implement the job planning stage
the idea is to built on top of the IterExplorer,
but have the "stack" of re-evaluation integrated
into a custom type, which exploits the static
node network structure to avoid heap allocations
solution idea: again use a builder function?
the template _Fun started as an internal helper
for function-closure, but seems to be of
general use. Thus move it into meta/function.hpp
(function-closure.hpp is heavyweight)
this enables expansion of a (functional) data structure
until exhaustion -- which is what we need to
build job functors by traversing and expanding
an arbitrarily nested job definition structure
the intention is to use this to simplify
generating render jobs based on the elaborated
dependency network of the render nodes. The key
challenge is to overcome the necessity to
store partially done evaluations as
continuation
the tricky part seems to be how to combine the
source iterators into a new monad instance, while
keeping this "Combinator" Strategy configurable
...just passes the compiler, while still lacking
even the generic implementation of joining
together the source iterators
The idea is to avoid building a data structure
for intermediary results, while still being able
to process a variably sized and arbitrary shaped
set of source data
without any trickery, we'll always get two indirections.
Thus, the decision is to turn the 2nd indirection
into a VTable call; this way, basically the JobClosure
also acts as job functor itself.
implemented as extension to the linear combinations.
I decided to use the same "always floor" rule
as employed for time quantisation. Moreover,
we don't support floating point, only rationals
funny enough this possible memory corruption
didn't happen in the unit test, because my
compiler optimised the additional int field
of class SubDummy, making it the same size
of the baseclass. Now matters should be safe.
this draft is based on
- Cehteh's draft for the scheduler
- my plannings about segmentation and JobTicket
it defines "Job" as a closure which can be invoked
from plain-C, using the information in the
job descriptor datastructure
also touches the question how to represent the job
descriptor datastructure. @Cehteh: I've just pasted
in your preliminary data struct definitinons
from the relevant mailing list discussions.
instead of (ab)using the Timings spect for a
runtime switch, better use the existing
MockInjector facility and thus turn the
mock engine mode into a global switch
I think it's smart to rather use ALSA directly instead of PortAudio.
ALSA is push AFAIK, and talking about it here at the hackspace, seems
like the better choice. It's a bit lower level, but anyway everything
speaks ALSA anyway. It's not like there's any reason to use PortAudio
at all. It's just an extra abstraction.
Coding for ALSA it'll also work with Pulseaudio and esd. Do people
really use other sound systems than Pulseaudio, esd or plain ALSA?
I can't think of it.
I really the idea about building a small tool first. I'll do that.
Also thought about making a small blikning cursor/text output, and
syncing a BEEP-sound to that, so that I can test around with throwing
in lots and lots of latency between "me" and the video, and try to
sync it anyway.
I should be able to read back from the sound card (or pulse audio
underneath, it will just work with alsa as the abstraction) how long
it takes for the bytes I'm pushing to reach the speakers, and do some
buffer tuning on that.
The goal is to make the build scrips more clear
at first sight. So move the main build targets
into a separate SConscript, to make them similar
to the tests, research and tools.
The final goal of this makeover is to reduce
the main SConstruct as much as possible
this turns out to be the typical usage scenario
for ScopedCollection: a manager object owning
the collection will populate it with specially
crafted components and invoke a member function
for creating the individual components.
This shortcut avoids using tr1::bind alltogether
