Update Graph.md

Graph.md

@@ -1,54 +1,212 @@
 # Graph
 
-## Pivot table UI
+Sqliteviz supports graph (network) visualisation. To build a graph, first run a query to get data.
+Then open the visualisation panel by clicking ![](./img/visualisation.svg)
+in any of the two-side toolbars and choose the graph mode by clicking ![](./img/graph.svg).
 
-Sqliteviz allows building pivot tables and visualizing them. To build a graph
+## Requirements for result set
-run a query to get data. Then open visualisation panel by clicking ![](./img/visualisation.svg)
-in any of the two side toolbars and choose a pivot mode by clicking ![](./img/graph.svg).
 
-A pivot visualisation has the following settings:
+To build a graph, a result set must follow the following requirements:
 
-- Columns – choose one or more column names from the result set. The values in
+- the first column must contain JSON documents (subsequent columns in the result set are ignored)
-  the chosen columns will be column names of the pivot table.
+- each JSON document has a key indicating whether the document represents a node/vertex (value `0`)
-- Rows – choose one or more column names from the result set. The values in the
+  or an edge (value `1`)
-  chosen columns will be row names of the pivot table.
+- each JSON document representing a node has a key with the node/vertex id
-- Order of columns and rows.
+- each JSON document representing an edge has a key with the edge source node/vertex and a key with
-- Aggregator and its arguments – a function which will be used for pivot cell
+  the edge target node/vertex
-  calculation. An aggregator can have from zero to two arguments. An aggregator
-  argument is one of the columns of the result set.
-- View – pivot table visualisation. It can be a table, a heatmap, a chart,
-  etc. See some examples of different views of the same pivot table below.
 
-![Fig. 1: Table](./img/Screenshot_pivot_table.png)
+That is what is the minimum required for a graph, but the JSON documents can have more fields used
+in graph styling (read [Graph styling](#graph-styling)).
 
-![Fig. 2: Heatmap](./img/Screenshot_pivot_heatmap.png)
+See also an example in [How to get a result set suitable for graph visualisation][1].
 
-![Fig. 3: Horizontal Stacked Bar Chart](./img/Screenshot_pivot_barchart.png)
+## Graph structure
 
-There are several built-in chart views for a pivot. But you can build your own
+Start building a graph by setting a mapping from your documents to node and edge properties.
-with `Custom chart` view (fig. 4).
+Go to `Structure` > `Graph` and set the following fields:
 
-![Fig. 4: Custom Chart](./img/Screenshot_pivot_custom_chart.png)
+- Object type – a field containing `0` for node documents and `1` for edge documents.
+- Node Id – a field containing the node identifier
+- Edge source – a field keeping a node identifier where the edge begins.
+- Edge target – a field keeping a node identifier where the edge ends.
 
-> **Note:**  You can switch to other pivot views and back to `Custom chart` –
+This is already enough to build a graph with default styling and circular layout.
-> all your custom chart settings will be remembered. But if you switch the
-> visualisation mode from pivot to any other mode, unsaved changes will be lost.
 
-You can save any visualisation as an image by clicking ![](./img/camera.svg).
+![Fig. 1: Graph structure settings](./img/Screenshot_graph_structure.png)
 
-## Pivot table SQL
+## Graph styling
 
-Pivot table (in the form of a result set) can be built on the SQL-level and,
+### General
-technically speaking, can be visualised as any other result set. Practically
-though there are a couple of challenges with that:
 
-1. Visualising a dataset of long/tall shape is much more convenient in Plotly
+Set a background color of the graph in `Style` > `General` panel.
-   (chart editor) rather than one of wide/fat shape.
-2. SQLite doesn't have a special constructs like `PIVOT` or `CROSSTAB` in
-   its SQL dialect.
 
-[How to build a pivot table in SQL(ite)][1] explores two options with static
+### Nodes
-(or beforehand-known) and dynamic columns.
 
-[1]: ../How-to-build-a-pivot-table-in-SQ-Lite
+There are the following settings in `Style` > `Nodes` panel:
 
+- Label - a field containing the node label. Note that if the graph has too many nodes or the node
+  size is too small, some labels can be visible only at a certain zoom level.
+- Label Color - a color of node labels
+- Size - set a node size. There are 3 modes of node sizing: constant, variable and calculated.
+
+  - `Constant` means that all nodes have the same given size.
+  - `Variable` allows you to choose a field where the node size is taken from.
+  - `Calculated` allows you to choose a method that is used to calculate the node size.
+
+    Currently, 3 methods are available: total node degree, degree by in-edges and
+    degree by out-edges.
+
+  For variable and calculated sizing it's also possible to set the scale, the minimum
+  size and the sizing mode - area or diameter. In the diameter mode the difference
+  between node sizes is more noticeable.
+
+- Color - set a node color. There are 3 modes of node color: constant, variable and calculated.
+
+  - `Constant` means that all nodes have the same color.
+  - `Variable` allows you to choose a field by which the color is determined.
+    With this option you can also choose if the color value should be taken directly or mapped to a
+    color palette.
+    
+    `Direct` mode means that in the JSON document representing the node, the value in the selected
+    field is used as a color. The color value in the JSON document can be set in different
+    ways:
+
+    1. As Hex, 8-digit (RGBA) Hex:
+
+       ```
+       #000
+       000
+       #369C
+       369C
+       #f0f0f6
+       f0f0f6
+       #f0f0f688
+       f0f0f688
+       ```
+
+    2. RGB, RGBA:
+
+       ```
+       // As a string:
+       rgb(255, 0, 0)
+       rgb 255 0 0
+       rgba(255, 0, 0, .5)
+
+       // As a nested structure:
+       { "r": 255, "g": 0, "b": 0 }
+       ```
+
+    3. HSL, HSLA:
+
+       ```
+       // As a string:
+       hsl(0, 100%, 50%)
+       hsla(0, 100%, 50%, .5)
+       hsl(0, 100%, 50%)
+       hsl 0 1.0 0.5
+
+       // As a nested structure:
+       { "h": 0, "s": 1, "l": .5 }
+       ```
+
+    4. HSV, HSVA:
+
+       ```
+       // As a string:
+       hsv(0, 100%, 100%)
+       hsva(0, 100%, 100%, .5)
+       hsv(0 100% 100%)
+       hsv 0 1 1
+
+       // As a nested structure:
+       { "h": 0, "s": 100, "v": 100 }
+       ```
+
+    5. Named colors:
+
+       Case insensitive names are accepted, using the list of [colors in the CSS
+       spec][2].
+
+       ```
+       RED
+       blanchedalmond
+       darkblue
+       ```
+
+    When `Map to` option is selected, the value by the selected field can be anything because it
+    won't be used directly as a color. In this case each distinct value will be mapped to a certain
+    color, so nodes with the same value will have the same color. Click on a color palette to open a
+    palette selector.
+
+  `Calculated` color mode allows to choose a method that will be used to determine a color.
+  Currently, 3 methods are available: total node degree, degree by in-edges and degree by
+  out-edges. You can also choose a color palette that will be used in a mapping of calculated
+  values into an actual color.
+
+- Color As - defines how color mapping should work - continuously or categorically.
+
+  Continuous mode is more suitable when the mapped values have a meaningful order. It looks more
+  informative with sequential palettes. In that case the lowest value corresponds to the first
+  color in the palette and the highest value - to the last color. The color of each intermediate
+  value reflects the position of that value in the range.
+
+  Categorical mode just uses the next color in the palette for each new distinct value.
+
+- Colorscale Direction - use the selected palette as is or reverse it.
+
+### Edges
+
+There are the following settings in `Style` > `Edges` panel:
+
+- Direction - show an edges with or without arrows
+- Label - a field containing an edge label. Note that if the graph has too many edges,
+  some labels can be visible only at a certain zoom level.
+- Label Color - a color of edge labels
+- Size - set an edge thickness. There are 2 modes of edge sizing: constant and variable.
+
+  - `Constant` means that all edges have the same thickness.
+  - `Variable` allows you to choose a field where the edge size is taken from.
+
+    For variable sizing it's also possible to set the scale and the minimum size.
+
+- Color - set an edge color. There are 2 modes of edge color: constant and variable. They work
+  similar to the node color modes.
+
+- Color As - defines how color mapping should work - continuously or categorically, similar to the
+  node color setting.
+
+- Colorscale Direction - use the selected palette as is or reverse it.
+
+### Layout
+
+#### Circular
+
+In this layout all nodes are just placed along a circle.
+
+#### Random
+
+This layout places nodes randomly for each seed value. The seed value allows you to restore the
+random layout you liked when you open the inquiry and run the query next time.
+
+![Fig. 2: Random layout](./img/Screenshot_graph_random.png)
+
+#### Circle pack
+
+Arranges nodes as a bubble chart according to specified attributes. You can choose multiple
+hierarchy attributes to group nodes.
+
+![Fig. 3: Circle pack layout](./img/Screenshot_graph_circle_pack.png)
+
+#### ForceAtlas2
+
+A continuous graph layout algorithm. Read more details about the algorithm and its settings in the
+[article][3]. The algorithm works iteratively. When you choose ForceAtlas2 layout or run the query,
+it is automatically run 50 iterations of the algorithm. You can change the amount of steps run
+automatically in `Initial Iterations`. You can also run and stop the algorithm manually by
+clicking `Start`/`Stop` button.
+
+![Fig. 3: ForceAtlas2 layout](./img/Screenshot_graph_force_atlas2.png)
+
+[1]: ./How-to-get-result-set-suitable-for-graph-visualisation
+[2]: https://www.w3.org/TR/css-color-4/#named-colors
+[3]: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0098679

How-to-get-result-set-suitable-for-graph-visualisation.md

@@ -0,0 +1,120 @@
+# How to get a result set suitable for graph visualisation
+
+There are some [requirements for result sets][1] if you want to make a graph.
+Here is an example of building a query that returns a result set appropriate for graph 
+visualisation.
+
+Let's say, you have 2 tables:
+
+1. `house`:
+
+   | name       | points |
+   | ---------- | ------ |
+   | Gryffindor | 100    |
+   | Hufflepuff | 90     |
+   | Ravenclaw  | 95     |
+   | Slytherin  | 80     |
+
+2. `student`:
+
+   | id | name           | house      |
+   | -- | -------------- | ---------- |
+   | 1  | Harry Potter   | Gryffindor |
+   | 2  | Ron Weasley    | Gryffindor |
+   | 3  | Draco Malfoy'  | Slytherin  |
+   | 4  | Luna Lovegood  | Ravenclaw  |
+   | 5  | Cedric Diggory | Hufflepuff |
+
+Each student belongs to a certain house. 
+Let's say you want to build a graph with houses and students as nodes, where each house is linked 
+with its students.
+
+We are going to use [json_object][2] function to form JSON documents. The result set should contain 
+both nodes and edges and we have to provide a field indicating if the document represents a node 
+(0) or and edge (1). Let's provide it as `object_type`:
+
+```sql
+SELECT json_object('object_type', 0)
+FROM house
+UNION ALL
+SELECT json_object('object_type', 0)
+FROM student
+UNION ALL
+SELECT json_object('object_type', 1)
+FROM student
+```
+
+Note that we included `student` table twice. That is because the table contains not only students 
+but also their relationship to houses. So the records from the first union of `student` are used 
+as nodes and from the second one - as edges.
+
+Then we need to provide an ID for each node. Let's put it in `node_id` field. The `node_id` value 
+for students is taken from `id` column and for houses - from `name`: 
+
+```sql
+SELECT json_object('object_type', 0, 'node_id', name)
+FROM house
+UNION ALL
+SELECT json_object('object_type', 0, 'node_id', id)
+FROM student
+UNION ALL
+SELECT json_object('object_type', 1)
+FROM student
+```
+
+Each edge document must provide a node id where the edge begins and where it ends. Let's put it in
+`source` and `target`:
+
+```sql
+SELECT json_object('object_type', 0, 'node_id', name)
+FROM house
+UNION ALL
+SELECT json_object('object_type', 0, 'node_id', id)
+FROM student
+UNION ALL
+SELECT json_object('object_type', 1, 'source', house, 'target', id)
+FROM student
+```
+
+Basically, that is enough to build a graph. But it is not very meaningful without labels.
+Also, it would be nice to distinguish house nodes from student nodes by color.
+Let's put additional fields `label` and `type` that can be used in graph styling.
+
+```sql
+SELECT json_object('object_type', 0, 'node_id', name, 'label', name, 'type', 'house') AS graph_object
+FROM house
+UNION ALL
+SELECT json_object('object_type', 0, 'node_id', id, 'label', name, 'type', 'student')
+FROM student
+UNION ALL
+SELECT json_object('object_type', 1, 'source', house, 'target', id)
+FROM student
+```
+
+Run the query, the result set will look like this:
+
+| graph_object                                                                   |
+| ------------------------------------------------------------------------------ |
+| {"object_type":0,"node_id":"Gryffindor","label":"Gryffindor","type":"house"}   |
+| {"object_type":0,"node_id":"Hufflepuff","label":"Hufflepuff","type":"house"}   |
+| {"object_type":0,"node_id":"Ravenclaw","label":"Ravenclaw","type":"house"}     |
+| {"object_type":0,"node_id":"Slytherin","label":"Slytherin","type":"house"}     |
+| {"object_type":0,"node_id":1,"label":"Harry Potter","type":"student"}          |
+| {"object_type":0,"node_id":2,"label":"Ron Weasley","type":"student"}           |
+| {"object_type":0,"node_id":3,"label":"Draco Malfoy","type":"student"}          |
+| {"object_type":0,"node_id":4,"label":"Luna Lovegood","type":"student"}         |
+| {"object_type":0,"node_id":5,"label":"Cedric Diggory","type":"student"}        |
+| {"object_type":1,"node_source":"Gryffindor","target":1}                        |
+| {"object_type":1,"node_source":"Gryffindor","target":2}                        |
+| {"object_type":1,"node_source":"Slytherin","target":3}                         |
+| {"object_type":1,"node_source":"Ravenclaw","target":4}                         |
+| {"object_type":1,"node_source":"Hufflepuff","target":5}                        |
+
+Now in the graph editor, we can set mapping of the result set documents into node and edge 
+properties, set graph styles and get the following visualisation:
+
+![Fig. 1: Graph visualisation example](./img/Screenshot_potter_example.png)
+
+
+[1]: ./graph#requirements-for-result-set
+[2]: https://sqlite.org/json1.html#jobj