aiSee: Graph Layout FAQs

How large of a graph can aiSee handle?

There are no special restrictions as to the maximum number of nodes, edges, subgraphs, or the depth of subgraph nesting. The largest graph we have ever seen drawn with aiSee contained over 1,000,000 nodes and well above 1,500,000 edges. We are used to working with graphs with 40,000 nested subgraphs (even on a slow 1.6GHz PC with only 1GB RAM).

How can I speed up the layout calculation for huge graphs?

Here are some suggestions:

Do not use splines, bitmap icons, or fish-eye views.
Use a fast layout algorithm (such as dfs or tree).
In a hierarchical layout, specify equal_y_dist: yes to prevent integer overflow.
Skip the crossing reduction phase 2 (GDL attribute crossing_phase2, aiSee 2 command line option -nocopt2).
Change the state of all unfolded subgraphs to boxed. The layout calculation of boxed (nested) subgraphs is much faster since the layout of each subgraph is calculated separately.
Hide parts of the graph, for example by

If the pendulum method takes too long (if you are using aiSee 2, this is indicated by m in the message window), limit the number of iterations for the crossing reduction (graph attribute pmax, aiSee 2 option -pmax).
If the straight-line fine-tuning phase takes too long (if you are using aiSee 2, this is indicated by S in the message window), limit its maximum iteration number (graph attribute smax, aiSee 2 option -smax).
When using force-directed layout, reduce the number of iterations performed in the simulated annealing part of the algorithm (graph attribute fdmax).
aiSee 2 only: switch to the fastest mode (command line option -fast). This will drastically reduce all iteration limits and may result in poor layout quality.
aiSee 2 only: set a time limit (-timelimit option).

aiSee 2 User Manual: reducing layout time

How can I visualize aiSee’s internal layout calculations?

If you are using aiSee 2, some internal layout calculations can be animated under the control of command-line options. For trees, -tanim is usually good choice. For force-directed layout, we recommend using -epanim. Example:

> aisee -epanim torus.gdl

How can I integrate aiSee graph layout in my Web page?

aiSee can produce various graphic formats directly, including SVG, PNG and BMP, which can be displayed by all modern browsers. For example,

aisee3 --export out.svg in.gdl

writes a Scalable Vector Graphics (SVG) file, which can be directly integrated in a Web page. This approach offers the best results. Another possibility is exporting to a bitmap format. For example,

aisee3 --export out.png in.gdl

writes a file in Portable Network Graphics (PNG) format.

aiSee can even produce a complete HTML document which contains an image mapped with hyperlinks and/or JavaScript functionality as specified in the GDL source of the graph. Example:

aisee3 --export out.png --export out.html in.gdl

The corresponding command lines for aiSee 2 are:

aisee -svgoutput out.svg in.gdl aisee -pngoutput out.png -lm 0px -tm 0px in.gdl aisee -htmloutput out.html -pngoutput out.png -lm 0px -tm 0px in.gdl

Why is clustering an “experimental feature”?

Subgraph layout is quite challenging by nature. aiSee includes an experimental implementation of clustered layouts which is not actively supported or maintained. Nevertheless, many aiSee users find this feature quite useful.

Does aiSee implement incremental layout?

No. However, you can use the edge attribute priority to emulate incremental layout to a certain extent. Assigning a high priority to the edges of the initial graph and a lower priority to the newly added ones would gain you some control over layout changes.