This commit adds a sample application that implements an
initial draft of the mappa API. In particular the abstraction
between the application and the mappa software is being kept
as simple as possible - even still it will probably change
a few times - so do not start using this API just yet!

This commit closes #80 initial mapping design.

Signed-off-by: Harry van Haaren <[email protected]>

Testing with valgrind to fix leaks + access issues, #81

Signed-off-by: Harry van Haaren <[email protected]>

See details in #81

Signed-off-by: Harry van Haaren <[email protected]>

This enables the app to add and remove multiple targets.
Internally they're stored for book-keeping, but not
event-lookup is implemented yet. See #81

Signed-off-by: Harry van Haaren <[email protected]>

This commit is a significant step towards POC of using Ctlra
look-up-table for event mapping. The software target events
are stored in a linear array sorted by type, so the lookup of
the target is just an array access. The function pointer is
called if one is set.

This almost completes #81.

Signed-off-by: Harry van Haaren <[email protected]>

Error check to avoid segfault on NULL m or lut,
refactor to make it easier to store a list of lut's

Signed-off-by: Harry van Haaren <[email protected]>

this allows multiple lut layers to be present in a single
device. It also gives a good abstraction for using the
userdata pointer for the ctlra dev callbacks, and a place
to store state like current-layer for a device.

Signed-off-by: Harry van Haaren <[email protected]>

Segfaulting due to incorrect linked-lists / target setup

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

Enable other types of ctlra inputs than slider, exposing it to
the external API. Need to think about if this is the right solution
to the interface for mapping

Signed-off-by: Harry van Haaren <[email protected]>

This allows the lut switch to target a specific
integer lut for each device.

Signed-off-by: Harry van Haaren <[email protected]>

This commit enables layer switching by using an internal
mappa callback function that takes a dev_t *. This allows
the mappa instance to changes its active_lut, which causes
different actions to be triggered by each input event.

This closes #81.

Signed-off-by: Harry van Haaren <[email protected]>

This tests the cross-fader being bound to something else
on layer 2 - to ensure that the functionality gets transposed
correctly.

Signed-off-by: Harry van Haaren <[email protected]>

This commit checks the ids of controls being bound,
failing with -EINVAL if the target is out of bounds
for the hardware device being mapped.

Signed-off-by: Harry van Haaren <[email protected]>

This commit strenghtens the error checking on events as compared
to the events that are advertised by the device's metadata. If
the event id is beyond the advertised range, it is ignored and
a print issued to notify - this is due to invalid event/metadata
pairing in the device code.

Signed-off-by: Harry van Haaren <[email protected]>

This commit is the first version of the feedback prototype.
It uses the token concept, along with a unique name per source
of feedback.

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

Mostly working - need to dig into last few pointer issues

Signed-off-by: Harry van Haaren <[email protected]>

This commit fixes the array issues mentioned in the previous
commit, and adds a new source callback function in the app.
The value returned by the two functions is now different, and
a simple "musher" allows mapping a float value to 0-6 lights.

Signed-off-by: Harry van Haaren <[email protected]>

General cleanup of prints and noisy code, added error check
for dev if there's no device available.

Signed-off-by: Harry van Haaren <[email protected]>

This commit refactors main to have a seperate function for
binding - this needs to be abstracted out of the app and into
the mappa library at some point.

Some valgrind mem leaks on close fixed by cleaning up sources.

Signed-off-by: Harry van Haaren <[email protected]>

This commit adds the tailq infrastructure to support multiple
devices. This allows more than one ctlra instance to be mapped.

Signed-off-by: Harry van Haaren <[email protected]>

The missing return was causing invalid reads in valgrind.

Signed-off-by: Harry van Haaren <[email protected]>

This commit adds a bunch of destroy() functions, which
free the linked-lists of targets, lut, devices etc.

Signed-off-by: Harry van Haaren <[email protected]>

This approach uses a "token" that the application can
give to each callback - think function parameters. This
is in addition to the function pointer itself.

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

This enables the token style variable size method.

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

This commit reworks the pointer setting for the targets,
now working as expected. Updated docs to state that the
target id is not to be relied on.

Signed-off-by: Harry van Haaren <[email protected]>

API to match standard void *, size paradigm, token passthrough
now working it seems.

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

This commit tests the token mechanism with a 32 byte
payload, memcmp()-ing it with known correct data.

Signed-off-by: Harry van Haaren <[email protected]>

This allows testing with a number of controllers that
don't have >=14 slider controls. Closes #85.

Signed-off-by: Harry van Haaren <[email protected]>

LEDs are lighting up as before - layer switching is up next.

Signed-off-by: Harry van Haaren <[email protected]>

Added a dynamic value based on slider input to set the
feedback led value.

Signed-off-by: Harry van Haaren <[email protected]>

Need to implement input layer switching with new APIs

Signed-off-by: Harry van Haaren <[email protected]>

Layer switching not working correctly yet

Signed-off-by: Harry van Haaren <[email protected]>

Limitation is that currently only one device can be switched.

Signed-off-by: Harry van Haaren <[email protected]>

This allows testing with even more controllers.

Signed-off-by: Harry van Haaren <[email protected]>

This commit adds an API to lookup a target_id from name.
This should complete the basics required to load a config
file, and map any controller to targets based on the file.

Basic layer switching is only supported now - in future
it must be expanded to more flexible methods.

Signed-off-by: Harry van Haaren <[email protected]>

Add some files for config reading. MIT license, simple and
easy to use .ini file reading.

Signed-off-by: Harry van Haaren <[email protected]>

This commit adds initial config file loading, creating
mappings based on the .ini config file passed in. This
allows any application to write the file, then trigger
a reload in mappa, which should be enough to allow any
app to map/expose everything.

Signed-off-by: Harry van Haaren <[email protected]>

Also for easier compiling in future to have a reference file.
This file format is not stable yet - this is just a v1 draft.

Signed-off-by: Harry van Haaren <[email protected]>

POC of source info from app being pushed back to the device,
lighting up LEDs.

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

This commit refactors the binding loading config file into the
mappa.c file instead of main.c, which is application code.
Also mutes a few more prints that snuck by.

Signed-off-by: Harry van Haaren <[email protected]>

This commit implements buttons from the config file, and
maps the button value to the layer to switch to. Simple
layer switching from layer 0 to 1 is enabled on buttons
0 and 1, and the feedback id's are swapped to showcase
the different layer mapping.

Signed-off-by: Harry van Haaren <[email protected]>

This commit refactors the header file for mappa, moving
all binding related functions into the _impl.h file. This
abstracts the host from any binding activities, and exposes
just the mappa_load_bindings() function.

Signed-off-by: Harry van Haaren <[email protected]>

Allows sources to be removed and de-allocated.

Signed-off-by: Harry van Haaren <[email protected]>

This allows external apps to be linked against Ctlra
and use Mappa features. Given Mappa has no external
dependencies it is always available if an applicaiton
wishes to use it. If not, Ctlra can be used as-is.

Signed-off-by: Harry van Haaren <[email protected]>

This commit enables C++ compilation and linking by adding
the extern "C" { code. Also fixes a hardcoded path, now uses
the passed-in filename.

Signed-off-by: Harry van Haaren <[email protected]>

This should fix the travis CI compile warnings/failures.

Signed-off-by: Harry van Haaren <[email protected]>

This commit adds dynamic log verbosity to MAPPA, allowing
the user to set level. Ctlra debug level is set based on
Mappa level, so the verbosities rise in parallel. This
can be over-ridden by using the CTLRA_DEBUG env var.

Signed-off-by: Harry van Haaren <[email protected]>

This fixes a use-after-free bug in previous code that read
the previously free-ed dev info. Now the dev is only freed
in the callback itself, which is the correct behaviour.

Signed-off-by: Harry van Haaren <[email protected]>

Previously a failure would cause other iters of that same loop
to not be handled, now it is. This code needs a refactor for
readability though.

Signed-off-by: Harry van Haaren <[email protected]>

This is working towards cleaner layer switching and more useable
APIs for layer switching.

Signed-off-by: Harry van Haaren <[email protected]>

This enables lut bindings being set by the config file
as the name will become the main identifier for each layer.

Signed-off-by: Harry van Haaren <[email protected]>

This commit changes the identifier from the lut_t to be
the name instead of the ID. As part of this change, some
APIs had to be updated to accept a const char* as parameter
instead of integer layer, which causes API/ABI break, so
rework of all apps etc expected.

Signed-off-by: Harry van Haaren <[email protected]>

This commit is a ~40% conversion towards using control names
as the config file items, instead of slider.0 and slider.1 etc.

This will improve the human-readable aspect of the mappa config
hugely, and also simplifies generation of bindings in a mapping
editor (no string<->int conversion between target and slider id).

Signed-off-by: Harry van Haaren <[email protected]>

This commit refactors a large amount of the config file loading,
scraping the item names from the Ctlra device metadata. The config
file is now human readable "key = value" pairs, with "key" being
a Ctlra device driver control name, and "value" being the target
string of the software to map to.

"Key" is defined by the code implemented in the Ctlra driver. The
names of the controls *MUST* refect the hardware device. If no
logical name exists, a self-descriptive must be used to identify
that singular control. Changing button names is *NOT* permitted,
as doing so approximates an ABI break for users' config files.

Signed-off-by: Harry van Haaren <[email protected]>

This commit prefixes the names being looked up in the config
file with the appropriate prefix, based on item type.

Signed-off-by: Harry van Haaren <[email protected]>

This TODO solution probably handles the dynamic target add
case well - users shouldn't notice that the "hot-plug" of the
app took place - thier controller should just light up with
the correct feedback.

Signed-off-by: Harry van Haaren <[email protected]>

This commit fixes an issue if iterating over the event_names
array, which would overflow the global array as it didn't have
the feedback item.

This commit also adds a compile-time check to ensure that these
arrays are the same size in future.

Signed-off-by: Harry van Haaren <[email protected]>

Detected using sanitiser=address, cleaned up by freeing
it in the appropriate dev_destroy() function.

Signed-off-by: Harry van Haaren <[email protected]>

Previously if there was no virtual or physical device attached
a segfault would occur as dev_t* was never checked.

Signed-off-by: Harry van Haaren <[email protected]>

This ensures that a name is passed in when creating a lut

Signed-off-by: Harry van Haaren <[email protected]>

This was freeing the lut_t * even though we didn't want it to,
which would later lead to segfaults / use-after-free.

Signed-off-by: Harry van Haaren <[email protected]>

The "mapping_count" item in a layer allows the human to indicate
the number of mappings that are created for this layer. This allows
mappa to check that exactly that number of mappings are made. If
fewer are made, warning/error prints are emitted to inform the
user that something has gone wrong.

WIP on active-on-load, which should enable activating layers once
the file is loaded.

Signed-off-by: Harry van Haaren <[email protected]>

This commit adds the concept of the active_lut_list, which contains
the currently active layers in the order that they should be handled.
The entries in this lut_list must be itereated and flattended into the
dev->active_lut in order for them to be used for event input lookup.

Signed-off-by: Harry van Haaren <[email protected]>

This commit uses the [mapping] layer_count field to get
the number of layers defined in the mapping, which is then
iterated to in the mapping lookups.

Signed-off-by: Harry van Haaren <[email protected]>

This commit implements lut_t flattening based on the order that
the layers are read from the config file. If a layer is active,
it is appended to the active_lut_list, which later gets flattened
into the active_lut.

Signed-off-by: Harry van Haaren <[email protected]>

This is a WIP commit but does prove the concept of the layer
active switching working with the sample .ini file.

This shows that the mapping can be achieved using the overlay
techinque, and that the button mapping to the layer-switch target
allows the user to program all aspects of multilayered mappings
using just the .ini config file, and some knowledge of the
controller and software in question.

Note that with virtual devices, a software developer can easily
assist the user with any issues that arise - actually the development
of this commit (and most of the previous mappa related ones) was
done entirely using virtual devices, away from the hardware. With
the hardware present, the cross-over was seamless to have working
input mappings with those devices.

Signed-off-by: Harry van Haaren <[email protected]>

This commit adds the demo config file to use with the mappa
branch, so others can test mappings easier. It is possible
create a virtual device in order test this mapping if you
don't have access to the Z1 hardware:

$ export CTLRA_VIRTUAL_VENDOR="Native Instruments"
$ export CTLRA_VIRTUAL_DEVICE="Kontrol Z1"

Now run ./examples/ctlra_mappa, and the middle "mode" button
will remap the "A" headphones que button (top left button)
between two different targets - NO application logic in
multiplexing, and NO user-programming - just a config file.

Signed-off-by: Harry van Haaren <[email protected]>

Remove some prints and add a TODO note on how to enable different
types of button actuation types in future.

Signed-off-by: Harry van Haaren <[email protected]>

This commit implements value scaling inside Mappa. The end
result is that an application can set a range for each target,
which Mappa then automatically range-scales and offsets any
float value mapped to it.

The end result is "transparent" range-scaling for users, where
a full movement of the hardware translates to full movement of
the software parameter (aka; basic expected experience).

However the value of having the target range exposed here is that
mappa can in future support "ranges" for each control, which can
be user defined, and the range numbers are not 0-1 based, but actual
real units (eg, remap lowpass filter min value from 0Hz to 110Hz, so
the sound is never 100% cut out).

Signed-off-by: Harry van Haaren <[email protected]>

This commit refactors the previous "load_bindings()" function,
replacing it with the more generic add_config_file(). This allows
the application to indicate config files that are present, and
gives mappa freedom to scan/load them as appropriate on hotplug
of devices.

All in all, this reduces complexity in the application (eg: no
need to implement a callback to get the correct mappa config for
a specific devices). It also allows future mappa to timestamp
loads of files, and auto-reload files etc.

Note that this commit breaks API/ABI of load_bindings()

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

Some %d %u mixups, some variable scope reductions.

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

This commit removes the mappa_check() function call
from the MAPPA_ERROR macro, as it *should* always print.
Code-checkers were throwing up all sorts of errors saying
that the limited data type comparison was invalid.

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

Note that although the code here is valgrind-clean, there
is an issue in .ini files being leaked.

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

This commit starts using the header-only TinyDir library
for cross-platform directory/file parsing.

Signed-off-by: Harry van Haaren <[email protected]>

Preparing to connect the callbacks in the foreach() config
function to the binding part of mappa. Need to refactor the
existing code quite a bit, so some breakage expected.

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

Also improve the error messages related to them.

Signed-off-by: Harry van Haaren <[email protected]>

Signed-off-by: Harry van Haaren <[email protected]>

This avoids trying to load config files before the hardware has
been scanned.

Signed-off-by: Harry van Haaren <[email protected]>

This allows mappa to provide better debug info, or possibly to
reload the file if required later (eg: on update).

Signed-off-by: Harry van Haaren <[email protected]>

The revision can be used to detect when we have added/removed
targets. This paves the way for "hotplug" of targets, and allowing
them to be removed/added and the mappings be picked up automatically.

Signed-off-by: Harry van Haaren <[email protected]>

Cleaning up output, and moving towards cleaner code.

Signed-off-by: Harry van Haaren <[email protected]>

This commit notifies users who *are* using the serial
option for thier scripts that the script applies only
to a specific device. Previously all users were notified
if thier script did not contain a serial code.

Signed-off-by: Harry van Haaren <[email protected]>

If a layer *says* it has a 0 mapping_count, don't print an
error if it also doesn't have any mappings.

Signed-off-by: Harry van Haaren <[email protected]>