Summarize property graph

[Dtenwolde]

Show statistics on the graph.

[Dtenwolde]

Similar to SUMMARIZE <table>, we can have a similar feature for summarizing a property graph or even just an edge/vertex table. This discussion is meant as an overview of what should be included as statistics and the exact syntax to be used.

This is different from the DESCRIBE <table> which contains more of an overview on the catalog information on the table. We support DESCRIBE PROPERTY GRAPH <pg>, which is similar.

For this feature, I think also we should do SUMMARIZE PROPERTY GRAPH <pg> and list statistics on all the element tables. Should this be limited to edge tables?

For reference, the output of SUMMARIZE person is the following:

⚫◗ summarize person;
┌─────────────┬─────────────┬─────────┬────────────────┬───────────────┬────────────────────┬────────────────────┬───────────────┬────────────────┬────────────────┬───────┬─────────────────┐
│ column_name │ column_type │   min   │      max       │ approx_unique │        avg         │        std         │      q25      │      q50       │      q75       │ count │ null_percentage │
│   varchar   │   varchar   │ varchar │    varchar     │     int64     │      varchar       │      varchar       │    varchar    │    varchar     │    varchar     │ int64 │  decimal(9,2)   │
├─────────────┼─────────────┼─────────┼────────────────┼───────────────┼────────────────────┼────────────────────┼───────────────┼────────────────┼────────────────┼───────┼─────────────────┤
│ id          │ BIGINT      │ 14      │ 37383395355368 │     11765     │ 18531368891241.254 │ 11363619254795.596 │ 8796093027141 │ 17723716063841 │ 28587302326733 │ 11000 │      0.00       │
└─────────────┴─────────────┴─────────┴────────────────┴───────────────┴────────────────────┴────────────────────┴───────────────┴────────────────┴────────────────┴───────┴─────────────────┘

So stats are:

• min
• max
• approx_unique
• avg
• std
• q25,50,75
• count
• null_percentage

For our statistics, it would be more related to graph specific statistics

• Number of vertices (select count(*) from vertex_table)
• Number of edges (select count(*) from edge_table)
• Number of connected components (can use the wcc udf)
• Largest connected component (can use the wcc udf)
• Average vertex degree
• Min / max degree

SELECT 
  average_degree: avg(count_), 
  min_degree: min(count_), 
  max_degree: max(count_)
FROM (
  SELECT src, count(*) count_ 
  FROM edge_table 
  GROUP BY src
);

• Graph density (E/(V*(V-1))

with v_count as (select v: count(*) from person), e_count as (select e: count(*) from person_knows_person) select graph_density: e/(v*(v-1)) from v_count, e_count;

• Number of isolated vertices
• Percentage of nodes with no in-edges / out-edges
• Degree distribution sketch (similar to q25, q50, q75)