{5} Assigned, Active Tickets by Owner (Full Description) (3 matches)

I propose two format modifications: the addition of edge ids and the use of separate attribute files.

Edge ids are an obvious necessity for separate attribute files, given the possibilities of parallel edges, so I shall focus on the latter.

Reasons

To reduce re-writing network files.

To make the addition of attributes easier for algorithm authors, particularly those using languages such as fortran that have fewer built-in string manipulation capabilities.

Supporting Infrastructure

For separate attribute files to work, it must be possible to use those attribute files in conjunction with the original network files. Due to the way the data manager works, this implies the need for several associated algorithms to enable sane workflows.

Merge Attributes

Select an NWB file and any number of attribute files (which will almost certainly be children of the NWB file), and this will generate a new NWB file with the attribute files included directly. This does involve rewriting a potentially large file, but since it is only done as necessary, and at user request, and can be done in a memory-efficient way, this is probably acceptable. The process should alert the user of possible conflicts (in attribute name) and errors (non-matching node or edge ids), and offer options for resolving those errors. It probably makes sense to offer override options when the algorithm is run, using input parameters, to enable remote usage and reduce tedium for experienced users. Note that having a variable-argument algorithm like this will require revision of the input_data algorithm property. There should probably also be an option to discard attributes in the original NWB file.

Extract Attribute

Given an NWB file, offer a list of the attributes (annotated with their association with nodes or edges) and extract the attribute chosen into an attributes file.

Convert Attribute to Table

Attributes have an obvious mapping to Tables. We should be able to support extensive useful operations on tables over time (top N, graph, et cetera) that become immediately useful on attributes.

Major Concerns

Unfortunately, there is no sane way to merge attributes with other network types; expressing this for the framework would involve a further modification of the in_data behavior.

Furthermore, this causes the user to have to become somewhat format-aware, which is unfortunate, as ideally format would not matter significantly. In practice this seems to be necessary, though. This will probably also necessitate a convert to NWB menu item (or a general forced conversion option that operates much like save or view).