feat: IPLD graph for explore page #672

olizilla · 2018-05-24T10:29:30Z

This PR adds a graph to the "explore" page to show the links from the current node and their target node type. It uses cytoscape and dagre layout.

CBOR node	Storybook

Introducing GRAPH CRUMB!

It also adds a "graph crumb" component, which shows the node containment and path boundaries that have been traversed along the user provided path.

This change required upgrading the logic for resolving info from them the path. Previously we passed the cid+path to dag.get and tried to render whatever value came back. As dag-cbor node values don't provide the same info as dag-pb nodes (no links or multihash properties) this pushed complexity on to the UI code of working out how to handle the differences. This PR fixes that by normalising data from dag.get into a common shape. The normaliseDagNode function currently handles dag-pb and dag-cbor but the plan is it can be extended to normalise all the things.

resolveIpldPath gives the additional info about the path by recursively calling dag.get from the root cid to find all the path boundaries (links that have to be traversed) to resolve the entire path. The normalised array of nodes and pathBoundaries that were crossed are made available to the the UI so the Graph crumb component can visualise it.

License: MIT
Signed-off-by: Oli Evans [email protected]

Uses cytoscape to display all the links from a given node License: MIT Signed-off-by: Oli Evans <[email protected]>

whyrusleeping · 2018-05-28T04:59:38Z

In the case where you have "favorites/{1,2,3}" it would be cool to represent "favorites" as its own node in the graph.

License: MIT Signed-off-by: Oli Evans <[email protected]>

olizilla · 2018-05-29T13:33:35Z

@whyrusleeping I'm thinking about the graph demonstrating links between IPLD nodes, and visually associating the link target nodes with the table of links. It's not obvious in these screenshots but the data explorer section shows the complete, possibly complex and deeply nested object tree within a single node.

My current plan is to improve the graph to show the IPLD node that a given path traverses, so you can get a feel for the path you took to get to given node, like:

...but I wasn't thinking about using it to show the internal structure of a CBOR node. Should I reconsider? Do you think it'd be more useful if it did show nodes that represent the internal CBOR structure as well?

whyrusleeping · 2018-05-29T15:49:06Z

@olizilla I think that showing internal structures (of every potential node type, not just cbor) would be very useful. cc @diasdavid

olizilla · 2018-05-29T21:08:19Z

@whyrusleeping @diasdavid please do add ideas for IPLD graph visualisations over here ipfs/ipfs-gui#11

License: MIT Signed-off-by: Oli Evans <[email protected]>

…link normalisation! License: MIT Signed-off-by: Oli Evans <[email protected]>

License: MIT Signed-off-by: Oli Evans <[email protected]>

Show the path boundaries and source nodes along a path in a neat breadcrumb style License: MIT Signed-off-by: Oli Evans <[email protected]>

License: MIT Signed-off-by: Oli Evans <[email protected]>

Where you explore into a sub path of a cbor node, we always show the info about the containing node, but show the exact path in the url and graph crumb. License: MIT Signed-off-by: Oli Evans <[email protected]>

…aph-cytoscape

olizilla · 2018-06-11T12:17:42Z

Sorry, this PR has grown much larger than I would have liked, but there were a bunch of details to resolve in it around node normalisation, path resolving, and handling paths that land somewhere in the middle of a cbor node.

olizilla · 2018-06-11T12:18:53Z

@lidel please can you take a look at this when you get a moment.

olizilla · 2018-06-11T12:31:45Z

The dag node normalisation logic tries to nudge things towards a more coherent naming scheme; normaliseDagNode returns objects like:

{
  cid: 'zHash1' // `cid` instead of `hash` or `multihash`
  type: 'dag-cbor' // or 'dag-pb'
  data: { /* the value of a dag-cbor node or the Data prop of a dag-pb */ }
  links: [
   {
      path: 'a/b/c' // in dag-cbor a link name may be a path, so standardise on `path`
      source: 'zHash1' // CID of node that contains this link.
      target: 'zHash2' // CID that the link points to.
  }
}

We map dag-cbor and dag-pb nodes returned from dag.get into this shape which then reduces the amount of conditional logic needed in the UI code. Adding the source cid to links is super useful as the link data can then be used in isolation from the parent node and still make sense, the source CID + path is a permanent-web alias for the target cid.

License: MIT Signed-off-by: Oli Evans <[email protected]>

olizilla · 2018-06-12T12:53:43Z

Am resolving the conflict now

…aph-cytoscape

lidel

Test CID that links from dag-cbor to dag-pb: zdpuAs8sJjcmsPUfB1bUViftCZ8usnvs2cXrPH6MDyT4zrvSs

That ↑ graph looks really slick and the way graph crumb provides context is 👌
Introducing unified abstraction format for different dag types makes sense, but I would avoid reusing path as a field name (made separate comment about it).

Below is a list of issues I've found while playing with it, but since this PR is only about adding a graph, I'm 👍 to merge it and address these things in separate PRs / issues.
Let me know if I should move below to dedicated issues.

Small potential bugs / edge cases that need additional handling:

When I open QmcPZGXSyaQSnDVEtJrgtp2EyxhENiBdpGa8pfs5rMBCWr I get this error in Console, and bit weird looking empty 'name' column. Perhaps name column should be skipped if none of items has value for it?
Opening QmZ5tQsSSdKS5Lez75KDfNxZ4usEYjHML5RBpPM6Hn2Z4n and clicking on data expander freezed my browser and triggered this warning 😱

UX stuffs:

CID and size columns overlap in narrow views
When user clicks on next node link or enters a new hash via input at the top, nothing happens until it is successfully fetched from the network.
- I strongly feel UI should immediately remove details of the old one node, update the CID at the top and display a "Fetching <CID> details from the swarm.." message to the user while CID details load in background.
- UI should be resilient to IPFS swarm hiccups. Right now unreachable CIDs will hang UI forever (example). I think there should be 5-10s timeout displaying an option to continue lookup (this time without timeout), or at least message informing user that lookup is in progress, as suggested in previous point.
- It should be possible to instantly abort lookup by entering a different hash.
  Right now nothing happens when this loads and I enter QmbWqxBEKC3P8tqsKc98xmWNzrzDtRLMiMPL8wBuTGsMnR and click "Explore"
Tiny things/ideas I've noticed while looking at it PR:
- Vertical alignment of the triangle expander in Data section is bit off, but it may be due to custom font setup on my Linux box.
- CID column gets cropped out when window size is narrow (example: cid, screenshot).
  May be a good idea to just add truncate class there, so that it gets text-overflow: ellipsis; sugar in edge cases. Looks a lot better than raw cut.
- When user enters hash into top "Explore" field, we should provide some visual feedback, eg. make button darker and change cursor to pointer to indicate it is ready to be clicked (right now it looks 'inactive' and I was surprised it worked :)) )
- Triangle expanders should have cursor: pointer; (pointer class) as additional hint they are actionable

lidel · 2018-06-12T14:13:28Z

src/explore/graph/IpldGraphCytoscape.js

+
+ makeNode ({target, path}, index) {
+ // TODO: pass link type info.
+ const type = target.startsWith('Q') ? 'dag-pb' : 'dag-cbor'


Yeah, someone will take CIDv0 of dag-pb and convert it to CIDv1 as a test and get weird results, like I did just a moment ago 🙃

lidel · 2018-06-12T14:20:16Z

src/lib/dag.js

+ */
+export function normaliseDagPbLinks (node) {
+ return node.links.map(({name, size, multihash}) => ({
+ path: name,


🚲 🏠 since path is actually either a path (cbor) or a name (pb), perhaps context or label would be a less confusing term?

olizilla · 2018-06-13T08:05:20Z

@lidel thanks for this, that's all really helpful. I strongly agree about making it super clear to the user what's happening and being robust when we find more exotic nodes. This PR gets things to a coherent point where it can work and is large, so I vote we merge it and tackle the issues raised in subsequent PRs.

Totally up for discussing, but right now I think path is a good name for the property as a name is a path of length 1. I see path as a superset of name, and we're using the term path for a "0 or more slash separated node traversal steps" ;-)

lidel

Makes sense, I have no strong feelings about that name so flipping to 👍

feat: IPLD graph for explore page

3794a72

Uses cytoscape to display all the links from a given node License: MIT Signed-off-by: Oli Evans <[email protected]>

hacdias added the revamp label May 24, 2018

wip: Add link index to graph.

f6ae314

License: MIT Signed-off-by: Oli Evans <[email protected]>

olizilla added 2 commits May 31, 2018 10:12

wip on graph crumb component

b076761

License: MIT Signed-off-by: Oli Evans <[email protected]>

wip: resolveIpldPath

9ed5545

License: MIT Signed-off-by: Oli Evans <[email protected]>

hacdias added the status/in-progress In progress label Jun 6, 2018

olizilla added 7 commits June 7, 2018 18:52

wip: resolveIpldPath now with canonicalPath, pathBoudaries, node and …

6c2f16b

…link normalisation! License: MIT Signed-off-by: Oli Evans <[email protected]>

wip: refactor in place, but dag.get with a path is not returning

a6008bb

License: MIT Signed-off-by: Oli Evans <[email protected]>

wip: work around remainderPath issue

199b15c

License: MIT Signed-off-by: Oli Evans <[email protected]>

wip: Add graph-crumb

798f5dc

Show the path boundaries and source nodes along a path in a neat breadcrumb style License: MIT Signed-off-by: Oli Evans <[email protected]>

wip: Handle paths that resolve part way through a cbor node

d111189

License: MIT Signed-off-by: Oli Evans <[email protected]>

wip: fix localPath handling

3d2a7bd

Where you explore into a sub path of a cbor node, we always show the info about the containing node, but show the exact path in the url and graph crumb. License: MIT Signed-off-by: Oli Evans <[email protected]>

Merge remote-tracking branch 'origin/revamp' into feat/revamp/ipld-gr…

fbf983d

…aph-cytoscape

olizilla requested a review from lidel June 11, 2018 12:15

wip: tidy up docs and remove console.logs

e60853f

License: MIT Signed-off-by: Oli Evans <[email protected]>

olizilla requested a review from hacdias June 12, 2018 10:43

Merge remote-tracking branch 'origin/revamp' into feat/revamp/ipld-gr…

d1e12eb

…aph-cytoscape

lidel requested changes Jun 12, 2018

View reviewed changes

lidel approved these changes Jun 13, 2018

View reviewed changes

olizilla merged commit caaadfb into revamp Jun 13, 2018

olizilla deleted the feat/revamp/ipld-graph-cytoscape branch June 13, 2018 10:11

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: IPLD graph for explore page #672

feat: IPLD graph for explore page #672

olizilla commented May 24, 2018 •

edited

Loading

whyrusleeping commented May 28, 2018

olizilla commented May 29, 2018

whyrusleeping commented May 29, 2018

olizilla commented May 29, 2018

olizilla commented Jun 11, 2018

olizilla commented Jun 11, 2018

olizilla commented Jun 11, 2018

olizilla commented Jun 12, 2018

lidel left a comment

lidel Jun 12, 2018

lidel Jun 12, 2018

olizilla commented Jun 13, 2018

lidel left a comment •

edited

Loading

feat: IPLD graph for explore page #672

feat: IPLD graph for explore page #672

Conversation

olizilla commented May 24, 2018 • edited Loading

Introducing GRAPH CRUMB!

whyrusleeping commented May 28, 2018

olizilla commented May 29, 2018

whyrusleeping commented May 29, 2018

olizilla commented May 29, 2018

olizilla commented Jun 11, 2018

olizilla commented Jun 11, 2018

olizilla commented Jun 11, 2018

olizilla commented Jun 12, 2018

lidel left a comment

Choose a reason for hiding this comment

lidel Jun 12, 2018

Choose a reason for hiding this comment

lidel Jun 12, 2018

Choose a reason for hiding this comment

olizilla commented Jun 13, 2018

lidel left a comment • edited Loading

Choose a reason for hiding this comment

olizilla commented May 24, 2018 •

edited

Loading

lidel left a comment •

edited

Loading