Force directed parameters seems to not have any impact

[EliotRagueneau]

I am trying to use your layout because it seems absolutely wonderful, but the different parameters related to the force-directed par of the layout seems to have 0 effects on the end result.
This is problematic because I have networks really divided, so I would like to increase gravity so that they are more "concentrated".

Did I do something wrong while using it, or are they simply not working?

[ugurdogrusoz]

@EliotRagueneau Gravity has an effect only within a certain range. A better solution to nicely packing your disconnected components is to use the packing facility supported by the latest CiSE in the "development branch" (not yet released). Please see the "packComponents" option.

Also, you should make sure that "randomize" is true (non-incremental layout, starts from random positions) when you're laying out your graph for the first time! Any succeeding layout where there are any incremental changes to the graph (new nodes/edges, removed nodes/edges), make sure "randomize" is false (incremental layout, respecting current positions).

Component packing works nicely with both incremental and non-incremental layout. However, you need to make sure that you use this "cytoscape-layout-utilities" extension properly as explained in the develop branch README and as applied in the develop branch demo.

[EliotRagueneau]

Thank you so much, it's much better using those options!

I believe you are aware of this, but the function for clusters (which is really useful!) doesn't work when you're using packComponents, I guess you will fix that later ^^ (In the meantime I calculate the clusters prior to running the layouts, it's working really well like that).

Develop version also seem to be slower than the master version, but I'm sure once released it's gonna have better performances ^^

Another small problem found is then with compound nodes, when the upper nodes are the clusters, it's a bit broken right now, but once again, I'm sure it will be fixed when you will release all those nice new features!

Maybe I could contribute to the project if you need some help, I would be happy to do so in my free time.

[EliotRagueneau]

Actually, after a bit more testing, I can tell you that the performance issue comes from the packComponents option with compound nodes:

	packComponents: false	packComponents: true
compound: false	2985ms	2786ms
compound: true	2767ms	11262ms

Might be hard to optimise that if it is coming from "cytoscape-layout-utilities" :/
Anyway, thank you very much for your answers and your help, my project will be much better thanks to you! For now, I'll just disable the compound option with the cise layout

[EliotRagueneau]

One last issue about the current dev version:

It has a strange behaviour with allowNodesInsideCircle:

As you can see, the node spacing inside clusters becomes irregular with my options

{
    name: 'cise',
    fit: true,
    padding: 30,
    allowNodesInsideCircle: true,
    randomize: true,
    packComponents: true
  }

In addition, performances get worse with packComponents

• 3801ms with only allowNodesInsideCircle = true
• 50531ms with both = true

Personnally I will simply not use allowNodesInsideCircle for now, but I just wanted to give you all the strange thing I found so far regarding your current develop version, just in case you were not aware of them.

[ugurdogrusoz]

Thanks for the thorough testing.

• CiSE does not do any clustering itself but assumes that they were calculated prior to layout and provided to it.
• CiSE cannot handle compound nodes/graphs. In fact, these are alternatives to each other: you either group nodes by compounds (use fCoSE for layout) or by clusters (use CiSE for layout).
• We are aware of the performance issues with packComponents (especially during incremental layout), haven't had a chance to work on this yet.
• If you allow nodes inside circles, CiSE will increase the separation amount for nodes on the circle when the nodes inside do not have enough room to settle.
• CiSE should not yield layouts like the last one you provided, where a cluster is inside another cluster! Could you provide us the JSON for this or perhaps put together a JSfiddle or something similar for us to reproduce it?

[EliotRagueneau]

Hi,

Sorry if I didn't express myself correctly about the clustering, I meant that you allow to put in clusters options either the whole clustering as an array, or a function that provides the cluster id for each node. Well, this function does not work with packComponents, so currently I have to use the array version, and I have to calculate this array before using the layout.

Here's the JSON used to build the data https://pastebin.com/dl/HRe94uJn, the clusters are based on the species (each species is a cluster). Hope this will help you

[herencalik]

Hello,

Thank you for providing us the data, however, the link doesn't seem to be working. Might there be a typo?

[EliotRagueneau]

Is this one working ? https://pastebin.com/HRe94uJn

[herencalik]

Yes, actually both of the links work. In fact, it seems restricted here so VPN solved the problem. Sorry for that.

[EliotRagueneau]

No worries, and thank you for being that responsive!

PS: You're doing a really great job, this layout is really awesome

[ugurdogrusoz]

@EliotRagueneau So, if you turn packing on and provide a function for cluster info, you have these overlapping clusters. But if you turn packing on and provide the cluster info as arrays, this does not happen, correct?

[EliotRagueneau]

No, function for cluster info simply doesn't work with packing. The only way to make it work is by providing cluster info as an array.

Then with the array, no matter the options, I often get clusters inside other clusters, even though it's more easy to spot when you have NodeInsideCircle, since it's complicating even more the layout.

For example, this screenshot has been produced with the following options:

{
    name: 'cise',
    fit: true,
    padding: 30,
    allowNodesInsideCircle: false,
    randomize: true,
    packComponents: true,
    clusters: [
      [
          "EBI-7916335",
          "EBI-7916387"
      ],
      [
          "EBI-1545818",
          "EBI-1545751",
          "EBI-541267",
          "EBI-1545845"
      ],
      [
          "EBI-1545762",
          "EBI-541307",
          "EBI-541321",
          "EBI-4426557",
          "EBI-541366",
          "EBI-4436798",
          "EBI-4424563",
          "EBI-1392127",
          "EBI-541381",
          "EBI-16890309",
          "EBI-16882503",
          "EBI-16946521",
          "EBI-16957889",
          "EBI-1810670",
          "EBI-16886194",
          "EBI-4455632",
          "EBI-16882482",
          "EBI-4431169",
          "EBI-16949315",
          "EBI-16945159",
          "EBI-16957271",
          "EBI-16956496",
          "EBI-16913889",
          "EBI-16882948",
          "EBI-16894943",
          "EBI-16893645",
          "EBI-16950521",
          "EBI-16953214",
          "EBI-4463452",
          "EBI-16753594",
          "EBI-541400",
          "EBI-2257898",
          "EBI-2257975",
          "EBI-16956097",
          "EBI-17066712",
          "EBI-16914575",
          "EBI-16929962",
          "EBI-16754362",
          "EBI-16901618",
          "EBI-16902452",
          "EBI-17080750",
          "EBI-16754134",
          "EBI-4450582",
          "EBI-541600",
          "EBI-541351",
          "EBI-4429530",
          "EBI-4430753",
          "EBI-4432023",
          "EBI-4430765",
          "EBI-4439121",
          "EBI-4424847",
          "EBI-4434651",
          "EBI-4424427",
          "EBI-16756869",
          "EBI-25506855",
          "EBI-25521272",
          "EBI-15681313",
          "EBI-4430993",
          "EBI-4424098",
          "EBI-4426585",
          "EBI-4426168",
          "EBI-4426504",
          "EBI-4427898",
          "EBI-1998719",
          "EBI-1100634",
          "EBI-963686",
          "EBI-1536925",
          "EBI-2119299",
          "EBI-1792431",
          "EBI-4424877",
          "EBI-4426901",
          "EBI-3133156",
          "EBI-531132",
          "EBI-2118417",
          "EBI-3387154",
          "EBI-4443010",
          "EBI-4465325",
          "EBI-4465331",
          "EBI-4430840",
          "EBI-4441246",
          "EBI-979475",
          "EBI-2297412",
          "EBI-4441249",
          "EBI-4424175",
          "EBI-4427398",
          "EBI-4444060",
          "EBI-4426649",
          "EBI-4427424",
          "EBI-4430035",
          "EBI-2292496",
          "EBI-4427395",
          "EBI-4425055",
          "EBI-4453767",
          "EBI-4472393",
          "EBI-4453099",
          "EBI-3133404",
          "EBI-4424482",
          "EBI-4424691",
          "EBI-4424503",
          "EBI-4424157",
          "EBI-307174",
          "EBI-4435148",
          "EBI-4441648",
          "EBI-4449554",
          "EBI-697501",
          "EBI-4446992",
          "EBI-4424568",
          "EBI-4466599",
          "EBI-4454548",
          "EBI-4437877",
          "EBI-4443654"
      ],
      [
          "EBI-15689882"
      ],
      [
          "EBI-1116104",
          "EBI-863137",
          "EBI-1127833",
          "EBI-545828",
          "EBI-1122941",
          "EBI-1118992",
          "EBI-9133821",
          "EBI-1120616",
          "EBI-562488",
          "EBI-541886",
          "EBI-551191",
          "EBI-561722",
          "EBI-1112775",
          "EBI-371008",
          "EBI-370963",
          "EBI-548960",
          "EBI-369221",
          "EBI-542092",
          "EBI-1131877",
          "EBI-545774",
          "EBI-1117994",
          "EBI-1124942",
          "EBI-1127454",
          "EBI-1113447",
          "EBI-1112893",
          "EBI-551481",
          "EBI-369487",
          "EBI-557681",
          "EBI-1117136",
          "EBI-557426",
          "EBI-553321",
          "EBI-369251",
          "EBI-543439",
          "EBI-543939",
          "EBI-544851",
          "EBI-543017",
          "EBI-545956",
          "EBI-543515"
      ],
      [
          "EBI-25567776"
      ],
      [
          "EBI-25601589",
          "EBI-5235612",
          "EBI-22758731",
          "EBI-11523345",
          "EBI-744081",
          "EBI-8099743",
          "EBI-11320284",
          "EBI-1051165",
          "EBI-347996",
          "EBI-6509505",
          "EBI-3905054",
          "EBI-79165",
          "EBI-727004",
          "EBI-2513582",
          "EBI-2548605",
          "EBI-1045507",
          "EBI-2800236",
          "EBI-741977",
          "EBI-723127",
          "EBI-357253",
          "EBI-21776319",
          "EBI-3905236",
          "EBI-355861",
          "EBI-12102178",
          "EBI-7205010",
          "EBI-746103",
          "EBI-77889",
          "EBI-3909537",
          "EBI-6929619",
          "EBI-720214",
          "EBI-747605",
          "EBI-1044712",
          "EBI-524064"
      ],
      [
          "EBI-15633769"
      ],
      [
          "EBI-21397",
          "EBI-21563",
          "EBI-2710",
          "EBI-29427",
          "EBI-26138",
          "EBI-7903",
          "EBI-13224",
          "EBI-2464632",
          "EBI-3809",
          "EBI-22211",
          "EBI-22178",
          "EBI-8648",
          "EBI-2332",
          "EBI-14344",
          "EBI-21449",
          "EBI-7699",
          "EBI-10420",
          "EBI-8603",
          "EBI-8627",
          "EBI-26482",
          "EBI-5222",
          "EBI-8666",
          "EBI-8591",
          "EBI-2989",
          "EBI-24159",
          "EBI-2344281",
          "EBI-28528",
          "EBI-16817",
          "EBI-2343154",
          "EBI-25172",
          "EBI-28706",
          "EBI-912262",
          "EBI-14166",
          "EBI-32965",
          "EBI-8394",
          "EBI-13275",
          "EBI-2447",
          "EBI-29516",
          "EBI-7218",
          "EBI-6475",
          "EBI-9890",
          "EBI-2218",
          "EBI-3027",
          "EBI-8659",
          "EBI-15864",
          "EBI-9182",
          "EBI-9186",
          "EBI-9195",
          "EBI-11476",
          "EBI-19530",
          "EBI-26411",
          "EBI-20212",
          "EBI-28914",
          "EBI-20589"
      ],
      [
          "EBI-21245803",
          "EBI-21245777"
      ],
      [
          "EBI-862771"
      ],
      [
          "EBI-3406532",
          "EBI-191805",
          "EBI-148693",
          "EBI-95916",
          "EBI-174123",
          "EBI-107815",
          "EBI-150802",
          "EBI-15120231",
          "EBI-82519",
          "EBI-186595",
          "EBI-497828",
          "EBI-174007",
          "EBI-88132",
          "EBI-15141584",
          "EBI-15135466",
          "EBI-15109101",
          "EBI-26696235",
          "EBI-15107521",
          "EBI-15106509",
          "EBI-15106535",
          "EBI-15107367",
          "EBI-85165",
          "EBI-15130484",
          "EBI-15107089",
          "EBI-131659",
          "EBI-190674",
          "EBI-167567",
          "EBI-149544",
          "EBI-180189",
          "EBI-151445",
          "EBI-176113",
          "EBI-151014",
          "EBI-101869",
          "EBI-6173284"
      ],
      [
          "EBI-11172266"
      ],
      [
          "EBI-9685715",
          "EBI-9685670"
      ],
      [
          "EBI-2799605"
      ],
      [
          "EBI-4319195"
      ]
   ]
}

[EliotRagueneau]

Hi, I just noticed another odd performace with packComponents:
When changing from a force layout to cise layout, it's taking about 3s.
When changing from a circular layout to cise layout: it's taking about 40s, most of it from setShifts4PackingComponents, himself due to 2 calls to i.packComponents, according to Chrome performance recorder. I don't know if this help :/

[ugurdogrusoz]

Overlap of components is due to layout utilities extension (see this issue).
We hope to improve the performance with both modifications to this extension (see this issue) and with improvements to packing (see this issue).
These improvements might take a while though. Will notify as they are done through this issue.

[hasanbalci]

@EliotRagueneau Function for cluster info should be working now with packing in the develop branch. #37

[ugurdogrusoz]

@EliotRagueneau we just released version 2.

[EliotRagueneau]

Thanks a lot !