Legend

# comments
* new section or null/unknown value or delimiter between the attribute name and the data type
() optional

Preserved Key Words (case sensitive)

• *Nodes
• *DirectedEdges
• *UndirectedEdges
• id
• source
• target
• weight
• *int
• *string
• *float

NWB Format

# Comments
*Nodes (total numbers of nodes)
id*int      label*string            (nodeAttribute1*data_type)                (nodeAttribute2*data_type) ...
...
*DirectedEdges     (total numbers of edges)
source*int         target*int               (edgeAttribute1*data_type)                (edgeAttribute2*data_type)
...
*UndirectedEdges   (total numbers of edges)
source*int         target*int               (edgeAttribute1*data_type)                (edgeAttribute2*data_type)
...

Example

# This is a paper-author network.
# Author labels are author names.
# Paper labels are titles.
# Paper weights indicate # citations.
*Nodes   4
id*int      label*string                weight*int             node_type*string  
1           "Joe Ann"                   0                      "author"
2           "John Smith"                0                      "author"
3           "Bio Today"                 8                      "paper"
4           "Physics Tomorrow"          15                     "paper"
*DirectedEdges     2
source*int              target*int           weight*float          edge_type*string
1                       3                    0.66                  "wrote"
4                       3                    0.78                  "paper-citation"

Detailed Specification

1. There are three types of section headers and they are case sensitive:
   • *Nodes
   • *DirectedEdges
   • *UndirectedEdges
2. There are three data types, string, integer, and floating point.
3. The *Nodes section header MUST NOT be omitted in the NWB file. The "total numbers of nodes" after *Nodes is optional, and should be separated only by arbitrary number of spaces and/or tabs.
4. The list of nodes under the section of *Nodes MUST NOT be omitted and MUST provide a complete list including the node id and node label. The data type of node ids must be integer and the value must be bigger than 0; in another word, the minimum node id is 1). The data type of node labels must be string.
5. One of *DirectedEdges section or *UndirectedEdges section is required. If desired, both may be used, but at least one must be. A graph with non-empty *DirectedEdges and *UndirectedEdges sections is a hybrid graph. "Total numbers of edges" is optional after *DirectedEdges or *UndirectEededges, and should be separated only by arbitrary number of spaces and/or tabs.
6. All edges must be listed, none are inferred.
7. All string values (not including the attribute names) must be surrounded by double quotation marks. " can not exist in the middle of a string surrounded by double quotation marks.
8. * is a preserved key character
   • It is a pre-fix of node or edge section headers.
   • It is used as a delimiter between an attribute name and the corresponding data type.
   • It also represents any null/unknown value
   • * in strings surrounded by double quotation marks should be treated as a regular character.
9. All integers must be represented by a positive or negative number with no decimal.
10. All floating point numbers must include a decimal. -1.23e5 is acceptable notation, as are values such as 102.5
11. id, source and target MUST be integers. label MUST be strings.
12. Arbitrary number of spaces and/or tabs can exist between two columns.
13. After each section header, the meaning of each column MUST be specified in the next line with the format of attribute_name*data_type. Any comment line or empty is not allowed between the line of section header and the attribute line.
14. Attribute names CANNOT contain spaces and/or * in the middle and cannot be quoted. The values of data_type can ONLY be int, string, and float. attribute_name*data_type MUST be lower case.
    Example2.nwb

    *Nodes 4 id*int label*string weight*float node_type*string 1 "Joe Ann" 0.66 "author" 2 "John Smith" 0 "author" 3 "Bio Today" 0.78 "paper" 4 "Physics Tomorrow" 1.0 "paper" *DirectedEdges 3 source*int target*int weight*int edge_type*string 1 3 1 "wrote" 4 3 15 "paper-citation" 2 3 1 "wrote"

15. All null/unknown values (data type could be string, or integer or floating point) are always represented by *. An empty string is "" (two double quotations marks)
16. Each comment line must start with # and comment lines must start from the beginning of a line and can not append to any node/edge section headers, attribute lines or node/edge lists.
    Example3.nwb

    *Nodes 4 id*int label*string weight*int node_type*string #the following node has an unknown string value. 1 * 0 "author" 2 "John Smith" 0 "author" 3 "Bio Today" 8 "paper" #the following node has an unknown integer value. 4 "Physics Tomorrow" * "paper" *DirectedEdges 3 source*int target*int weight*float edge_type*string 1 3 0.66 "wrote" 4 3 0.78 "paper-citation" 2 3 1.0 "wrote"

17. Separate node and edge attribute sections will continue to NOT be supported.

Acknowledgements

NWB File Format is defined by the NWB Development Team