document() recompiles package source code, which requires a NAMESPACE

[cboettig]

Attempting to write a package wrapper around some existing source code. For reasons that are not obvious to me, calling devtools::document() starts the package recompiling. Moreover, this call fails if the package does not have a NAMESPACE file. If I start the package with a manually created NAMESPACE (even a blank one), then devtools::document() runs, recompiling everything, but then refuses to update my NAMESPACE file. Thus it appears impossible to have devtools create a NAMESPACE for the package.

Of course I can hack around this by starting with a NAMESPACE file with the comment

# Generated by roxygen2: do not edit by hand

but this feels wrong. Is there a more elegant work-around out of this catch-22? (Why does document() need to recompile anyhow?) Am I missing something?

[hadley]

Roxygen is a run-time documentation system - it inspects the objects in the package environment, not just the source code. That means it needs to load_all() the package, which obviously it can't do without compiling any C code.

Would you mind briefly outlining a series of steps that gets you to this point? Normally the NAMESPACE doesn't matter when compiling code.

[cboettig]

Apologies for not having a good minimal example, this could be user error from a naive attempt to wrap a C program. Anyway, if you clone https://github.com/cboettig/rappl/tree/testing and run devtools::document() you should see the error about not having a NAMESPACE.

Here's the run log I see:

devtools::document()
Updating rappl documentation
Loading rappl
Re-compiling rappl
'/usr/lib/R/bin/R' --no-site-file --no-environ --no-save --no-restore CMD  \
  INSTALL '/home/rstudio/rappl'  \
  --library='/tmp/RtmpDrJUwI/devtools_install_17fbdfe81' --no-R --no-data  \
  --no-help --no-demo --no-inst --no-docs --no-exec --no-multiarch  \
  --no-test-load --preclean 

* installing *source* package ‘rappl’ ...
rm -f ./PolicyGraph/PolicyGraph.o ./PolicyGraph/PolicyGraphGenerator.o ./Evaluator/EvaluationEngine.o ./Evaluator/EvaluatorSampleEngine.o ./Simulator/SimulationEngine.o ./MathLib/DenseVector.o ./MathLib/MathLib.o ./MathLib/SparseMatrix.o ./MathLib/SparseVector.o ./Algorithms/SARSOP/BinManager.o ./Algorithms/SARSOP/BinManagerSet.o ./Algorithms/SARSOP/Sample.o ./Algorithms/SARSOP/SampleBP.o ./Algorithms/SARSOP/SARSOP.o ./Algorithms/SARSOP/SARSOPPrune.o ./Models/MOMDP/BeliefTransitionMOMDP.o ./Models/MOMDP/ObservationProbabilities.o ./Models/MOMDP/VariableRelation.o ./Models/MOMDP/BooleanVariable.o ./Models/MOMDP/StateTransitionXY.o ./Models/MOMDP/StateTransitionY.o ./Models/MOMDP/MOMDPLite.o ./Models/MOMDP/MOMDP.o ./Models/MOMDP/VariableValue.o ./Models/MOMDP/Variable.o ./Models/MOMDP/StateTransitionXXpY.o ./Models/MOMDP/IVariable.o ./Models/MOMDP/VariableCombined.o ./Models/MOMDP/Rewards.o ./Models/MOMDP/StateTransitionX.o ./Models/MOMDP/BeliefTransitionMOMDPLite.o ./Models/MOMDP/IVariableValue.o ./OfflineSolver/GlobalResource.o ./OfflineSolver/solverUtils.o ./Bounds/AlphaPlane.o ./Bounds/AlphaPlanePool.o ./Bounds/AlphaPlanePoolSet.o ./Bounds/AlphaVectorPolicy.o ./Bounds/BackupAlphaPlaneMOMDP.o ./Bounds/BackupAlphaPlaneMOMDPLite.o ./Bounds/BackupBeliefValuePairMOMDP.o ./Bounds/BackupBeliefValuePairMOMDPLite.o ./Bounds/BeliefValuePair.o ./Bounds/BeliefValuePairPool.o ./Bounds/BeliefValuePairPoolSet.o ./Bounds/BlindLBInitializer.o ./Bounds/FastInfUBInitializer.o ./Bounds/FullObsUBInitializer.o ./Bounds/PruneAlphaPlane.o ./Bounds/PruneBeliefValuePair.o ./Bounds/xml_parse_lib.o ./Core/Actions.o ./Core/VariableContainer.o ./Core/Belief.o ./Core/BeliefCache.o ./Core/BeliefException.o ./Core/BeliefForest.o ./Core/BeliefTreeNode.o ./Core/BeliefWithState.o ./Core/MObject.o ./Core/Observations.o ./Core/States.o ./Core/UniqueBeliefHeap.o ./Parser/Cassandra/Parser.o ./Parser/Cassandra/POMDP.o ./Parser/Cassandra/pomdpCassandraWrapper.o ./Parser/ParserSelector.o ./Parser/POMDPX/FactoredPomdp.o ./Parser/POMDPX/Function.o ./Parser/POMDPX/ObsAct.o ./Parser/POMDPX/PreCEntry.o ./Parser/POMDPX/PreCMatrix.o ./Parser/POMDPX/SparseEntry.o ./Parser/POMDPX/SparseTable.o ./Parser/POMDPX/State.o ./Parser/POMDPX/StateObsAct.o ./Parser/POMDPX/tinystr.o ./Parser/POMDPX/tinyxml.o ./Parser/POMDPX/tinyxmlerror.o ./Parser/POMDPX/tinyxmlparser.o ./Utils/InfoLog.o ./Utils/md5.o ./Utils/md5wrapper.o ./Utils/CPMemUtils.o ./Utils/StatsCollector.o ./Utils/SimulationRewardCollector.o ./Parser/Cassandra/include/pomdp_spec.tab.o ./Parser/Cassandra/include/pomdp_spec.yy.o ./miniposix/getopt.o ./miniposix/getopt1.o ./miniposix/getopt_init.o ./Parser/Cassandra/decision-tree.o ./Parser/Cassandra/imm-reward.o ./Parser/Cassandra/mdpCassandra.o ./Parser/Cassandra/parse_err.o ./Parser/Cassandra/parse_hash.o ./Parser/Cassandra/sparse-matrix.o ./Controller/Controller.o   ./OfflineSolver/solver.o  ./Simulator/Simulator.o  ./Evaluator/Evaluator.o  ./PolicyGraph/PolicyGraph.o  ./PomdpConvertor/convertor.o  ./Controller/testControllerTiger.o 
rm -f libappl.a pomdpsol pomdpsim pomdpeval polgraph pomdpconvert testControllerTiger
ERROR: a 'NAMESPACE' file is required
* removing ‘/tmp/RtmpDrJUwI/devtools_install_17fbdfe81/rappl’
Error: Command failed (1)
> 

This seems to be related to the fact that I am using the Makefile from the original C program, which I understand isn't the right thing to do, so perhaps this isn't a problem once I figure out how to write the proper Makevars file in place of it, so perhaps this isn't a real issue. (I have a deal to learn still on that front). I was just surprised by the error.

[hadley]

How did you create the original package? Most ways should give you a dummy/empty NAMESPACE.

[cboettig]

I created it with devtools::setup() (actually copied a template created with setup()). This created a dummy NAMESPACE (that does not have the #Generated by roxygen2: header), and so Roxygen was refusing to overwrite that. So I deleted the NAMESPACE and wound up with the above error.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/