make install-models: delegate to ocrd resmgr

[kba]

Once OCR-D/core#559 is merged.

[bertsky]

I recommend a more radical change: Instead of unconditionally delegating install-models to install-models-ocropus, install-models-calamari etc, we should make these subtargets depend on whether the respective module is in the currently active OCRD_MODULES, and encapsulate each such rule near the other rules for the modules.

Thus, instead of …

ifneq ($(findstring ocrd_cis, $(OCRD_MODULES)),)
OCRD_EXECUTABLES += $(OCRD_CIS)
...

…we would have…

ifneq ($(findstring ocrd_cis, $(OCRD_MODULES)),)
install-models: install-models-cis
install-models-cis: $(BIN)/ocrd
	. $(ACTIVATE_VENV) && ocrd resmgr download ocrd-cis-ocropy-recognize '*'
OCRD_EXECUTABLES += $(OCRD_CIS)
...

Also, we still need to adapt Dockerfile:

ENV XDG_DATA_HOME /usr/local/share
VOLUME $XDG_DATA_HOME/ocrd-resources

(perhaps also HOME itself)

[kba]

Having thought about it some more, I agree that install-models-* should depend on OCRD_MODULES to ensure that only the models are installed for those processors the user requested (by having it in OCRD_MODULES)

[kba]

we should make these subtargets depend on whether the respective module is in the currently active OCRD_MODULES

fb4ee26

Also, we still need to adapt Dockerfile:

2fdce0a

[bertsky]

Splendid – thanks!

Perhaps we could offer more of these install-models-* rules in order to allow a quick installation without studying all processors and available models? (Currently, I would still have to ocrd resmgr download PROCESSOR '*' selectively for each processor I might want to use.)

[kba]

Perhaps we could offer more of these install-models-*

384ee6d I added targets for all projects that we have in core's resource_list.yml.

You could always do the installation and then ocrd resmgr download '*' to download all models at once later on.