Due to a copy-paste error, gray-scale plotting misbehaved. This is fixed
by setting proper names on the correct vectors.

Thanks to @ecklbarb for pointing this out.

Signed-off-by: Claus Hunsen <[email protected]>

When plotting networks using the default layout of the package 'ggraph'
(i.e., 'igraph::layout_nicely'), we experienced segfaults in R.
Therefore, we change the default layout to
'igraph::layout.kamada.kawai'.

This essentially fixes #109.

Thanks to @hechtlC for experiencing the problems and @bockthom for
helping to find a solution.

Signed-off-by: Claus Hunsen <[email protected]>

As we always match motifs on the vertex attribute 'type', the existing
parameter to the function 'motifs.search.in.network' is removed.

The test suite is adapted accordingly. Additionally, the documentation
is updated.

Signed-off-by: Claus Hunsen <[email protected]>

The vertex and edge types as defined in the networks module
('TYPE.AUTHOR', 'TYPE.ARTIFACT', 'TYPE.EDGES.INTRA', and
'TYPE.EDGES.INTER') are defined as magic numbers (historically coming
from vector indices). We refactor them to be of type 'character'. This
way, we are able to actually "read" the vertex and edge attributes
without any lookup from the network objects we construct.

The constants are now defined as follows:
TYPE.AUTHOR = "Author"
TYPE.ARTIFACT = "Artifact"
TYPE.EDGES.INTRA = "Unipartite"
TYPE.EDGES.INTER = "Bipartite"

As the motif identification works on numerics for the vertex encoding
(see 'igraph::subgraph_isomorphisms', method 'vf2'), the corresponding
code is adapted to transform the (now) character constants reliably back
to numerics. For this, we introduce the mapping 'MOTIF.TYPE.MAPPING' and
the function 'get.vertex.types.as.numeric'.

As we manually translate vertex and edge types to strings for display in
the plotting module, the module gets appropriate fixes and clean-ups.
Here, the legend name for author vertices is not 'TYPE.AUTHOR', but
"Developer". The behavior can be overwritten via setting
'PLOT.VERTEX.TYPE.AUTHOR' (and 'PLOT.VERTEX.TYPE.ARTIFACT',
respectively) appropriately.

This fixes #110.
Thanks to @bockthom for his numerous remarks on this change.

Signed-off-by: Claus Hunsen <[email protected]>

Signed-off-by: Claus Hunsen <[email protected]>

In some places, we still use the word "developer" in our codebase,
although we state in the contribution guide that we only use "author".
This is fixed by replacing the occurrences properly.

Thanks to @bockthom for pointing this out.

Signed-off-by: Claus Hunsen <[email protected]>

Signed-off-by: Claus Hunsen <[email protected]>

Various minor and major fixes and adjustments

Reviewed-by: Thomas Bock <[email protected]>

Don't add lists containing one element as vertex attributes any more, but
add just single vector elements. If a vertex attribute can have multiple
values per vertex (e.g., active ranges), then we need to wrap the
vertex-attribute vector into an individual list per vertex.

This has one side effect: Vertex attributes where simply are dates are not
saved in POSIXct format any more, but as UNIX timestamps (similar to edge
attributes).

This fixes #112.

In addition, all the corresponding tests are adjusted to the new behavior.

Signed-off-by: Thomas Bock <[email protected]>

The function 'get.date.from.string' does currently not handle time
zones, that is, it expects every input to be in UTC, even if there is
another time-zone suffix in the input.

To make the user aware of that, add a notice to this function and add an
additional test which makes sure that the time-zone suffix is, indeed,
ignored.

Signed-off-by: Thomas Bock <[email protected]>

Signed-off-by: Thomas Bock <[email protected]>

Just some small changes on comments.

Signed-off-by: Thomas Bock <[email protected]>

Fix lists of vertex attributes

Reviewed-by: Claus Hunsen <[email protected]>

Signed-off-by: Claus Hunsen <[email protected]>