[stackcollapse]

# How to guarantee unique ID's in SVG generated? **URL:** \ **Category:** Help\ **Created:** [October 4, 2024, 9:54pm UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473 "2024-10-04T21:54:30Z")\ **Posts on this page:** 14\ **Page:** 1

**Author:** ![hbmartin](https://sea2.discourse-cdn.com/graphviz/user_avatar/forum.graphviz.org/hbmartin/32/1635_2.png) [@hbmartin](https://forum.graphviz.org/u/hbmartin)\ **Post date:** [October 4, 2024, 9:54pm UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/1 "2024-10-04T21:54:30Z")

Is there a way to require that all of the `id`s in generated SVG are unique without manually specifying them? SVG spec requires that id’s in a document are unique, which is the behavior I expect based on [id | Graphviz](https://graphviz.org/docs/attrs/id/) but this doesn’t happen in practice. e.g. If I render the `pprof` example [pprof CPU Profile | Graphviz](https://graphviz.org/Gallery/directed/pprof.html) it has multiple `` tags with identical id’s ---

**Author:** ![smattr](https://sea2.discourse-cdn.com/graphviz/user_avatar/forum.graphviz.org/smattr/32/85_2.png) [@smattr](https://forum.graphviz.org/u/smattr)\ **Post date:** [October 5, 2024, 12:35am UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/2 "2024-10-05T00:35:40Z")

I guess the duplicates you’re pointing out are `node1` and `a_node1`? SVG content for anyone following along at home: ```xml ``` Not sure what the expected behavior is for a subgraph in this position. ---

**Author:** ![steveroush](https://avatars.discourse-cdn.com/v4/letter/s/a9adbd/32.png) [@steveroush](https://forum.graphviz.org/u/steveroush)\ **Post date:** [October 5, 2024, 1:22am UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/3 "2024-10-05T01:22:23Z")

I think it is a bug. The documentation for **id** says: ```auto If no id attribute is provided, then a unique internal id is used. However, this value is unpredictable by the graph writer. ``` I don’t believe that _node1_ was a _unique internal id_. However, there is an easy work-around - add an **id** to the one node that does not already have an **id** and all works as desired - no duplicates. ```auto subgraph cluster_L { "File: [stackcollapse]" [id=XYZ shape=box fontsize=16 label="File: [stackcollapse]\l\lShowing nodes accounting for 380, 90.48% of 420 total\lDropped 120 nodes (cum <= 2)\lShowing top 20 nodes out of 110\l\lSee https://git.io/JfYMW for how to read the graph\l" tooltip="[stackcollapse]"] } ``` ---

**Author:** ![steveroush](https://avatars.discourse-cdn.com/v4/letter/s/a9adbd/32.png) [@steveroush](https://forum.graphviz.org/u/steveroush)\ **Post date:** [October 5, 2024, 4:27am UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/4 "2024-10-05T04:27:58Z")

Whoops, not a bug! _node1_ is a unique internal ID. Below is a reduced version, but with all the nodes and IDs renamed. **Dot** still produces _node1_ as the ID for the one unnamed node. The moral of the story is add IDs to _all_ your nodes or use some ID naming scheme other than _nodeN_. ```auto digraph "[stackcollapse]" { node [style=filled fillcolor="#f8f8f8"] subgraph cluster_L { "File: [stackcollapse]" [shape=box fontsize=16 label="File: [stackcollapse]\l\lShowing nodes accounting for 380, 90.48% of 420 total\lDropped 120 nodes (cum <= 2)\lShowing top 20 nodes out of 110\l\lSee https://git.io/JfYMW for how to read the graph\l" tooltip="[stackcollapse]"] } N2 [label="deflate\n62 (14.76%)\nof 384 (91.43%)" id="FLOWER2" fontsize=18 shape=box tooltip="deflate (384)" color="#b20400" fillcolor="#edd6d5"] N3 [label="gzip\n0 of 409 (97.38%)" id="FLOWER3" fontsize=8 shape=box tooltip="gzip (409)" color="#b20100" fillcolor="#edd5d5"] N2 -> N3 [label=" 1" color="#b2b2b0" tooltip="zip -> flush_block (1)" labeltooltip="zip -> flush_block (1)"] } ``` ---

**Author:** ![scnorth](https://sea2.discourse-cdn.com/graphviz/user_avatar/forum.graphviz.org/scnorth/32/89_2.png) [@scnorth](https://forum.graphviz.org/u/scnorth)\ **Post date:** [October 5, 2024, 12:05pm UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/5 "2024-10-05T12:05:33Z")

Hmm. Not sure we thought about building a dictionary of object IDs like that, to avoid collisions when an input graph mixes explicit SVG IDs and automatically generated IDs. There’s various places where ad-hoc SVG id generation happens in [gvrender\_core\_svg.c](https://gitlab.com/graphviz/graphviz/-/blob/main/plugin/core/gvrender_core_svg.c) so you can see what is happening. It’s tempting to “improve” this, but wonder if it could break existing downstream processors, too. ---

**Author:** ![hbmartin](https://sea2.discourse-cdn.com/graphviz/user_avatar/forum.graphviz.org/hbmartin/32/1635_2.png) [@hbmartin](https://forum.graphviz.org/u/hbmartin)\ **Post date:** [October 5, 2024, 3:07pm UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/6 "2024-10-05T15:07:17Z")

@scnorth Thank you for the insight. I suppose per Hyrum’s Law there could be some downstream breakage but I’m having a hard time imaging where this would be explicitly desirable. Are there any particular processors you have in mind as a concern? Perhaps a smaller incremental improvement could be a runtime warning that a graph author should not use node\* id’s (or if so should set id’s on all nodes). However current behavior does break my use case ([graphviz2drawio](https://github.com/hbmartin/graphviz2drawio)) and is contrary to svg spec so I’m interested in contributing an id generation change if that’s something that would be merged. ---

**Author:** ![mark](https://sea2.discourse-cdn.com/graphviz/user_avatar/forum.graphviz.org/mark/32/84_2.png) [@mark](https://forum.graphviz.org/u/mark)\ **Post date:** [October 6, 2024, 4:49am UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/7 "2024-10-06T04:49:15Z")

Yeah, we have probably millions of graphs out there using Graphviz; Hyrum’s law is always a bit of a worry. For particular processors, I think some javascript graphviz editors take the SVG output and parse the IDs to turn them back into which input nodes should be updated when you click on a rendered node. e.g. [https://magjac.com/graphviz-visual-editor/](https://magjac.com/graphviz-visual-editor/), though I’m not totally sure if they use the ID. ---

**Author:** ![hbmartin](https://sea2.discourse-cdn.com/graphviz/user_avatar/forum.graphviz.org/hbmartin/32/1635_2.png) [@hbmartin](https://forum.graphviz.org/u/hbmartin)\ **Post date:** [October 6, 2024, 3:00pm UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/8 "2024-10-06T15:00:08Z")

@mark I don’t think that’s right, the gve editor is keyed on node name / title. It also has the same problem when exporting SVG of the pprof graph. ---

**Author:** ![magjac](https://sea2.discourse-cdn.com/graphviz/user_avatar/forum.graphviz.org/magjac/32/16_2.png) [@magjac](https://forum.graphviz.org/u/magjac)\ **Post date:** [October 6, 2024, 7:28pm UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/9 "2024-10-06T19:28:25Z")

@hbmartin You are correct. The correct terminology according to the [DOT language](https://graphviz.org/doc/info/lang.html) is however not _node name_ but _node\_id_ (not to be confused with the node attribute [_id_](https://graphviz.org/docs/attrs/id/)) ---

**Author:** ![smattr](https://sea2.discourse-cdn.com/graphviz/user_avatar/forum.graphviz.org/smattr/32/85_2.png) [@smattr](https://forum.graphviz.org/u/smattr)\ **Post date:** [October 7, 2024, 6:02am UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/10 "2024-10-07T06:02:03Z")

I’m inclined to agree with @hbmartin. Graphviz should detect this case and skip the colliding IDs when generating its own ones. I don’t think we can say “don’t mix explicit and implicit IDs” or “use a different naming scheme” when an example on our own website violates this. ---

**Author:** ![scnorth](https://sea2.discourse-cdn.com/graphviz/user_avatar/forum.graphviz.org/scnorth/32/89_2.png) [@scnorth](https://forum.graphviz.org/u/scnorth)\ **Post date:** [October 8, 2024, 3:46pm UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/11 "2024-10-08T15:46:40Z")

Are you proposing to just bite the bullet and maintain a dictionary of user-defined IDs so we can avoid collisions? (The typical coding style of graphviz would be to cheat a little and use a weird string prefix for system-generated IDs and hope for the best…) though what you are asking seems not that difficult. Also, just to make sure, these SVG IDs have nothing to do with node names (that are mapped into internal IDs), edge keys, etc. at the cgraph level. ---

**Author:** ![smattr](https://sea2.discourse-cdn.com/graphviz/user_avatar/forum.graphviz.org/smattr/32/85_2.png) [@smattr](https://forum.graphviz.org/u/smattr)\ **Post date:** [October 9, 2024, 7:09am UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/12 "2024-10-09T07:09:52Z")

I hadn’t yet thought about the “how”. I was still stuck on the “what”. > The typical coding style of graphviz would be to cheat a little and use a weird string prefix for system-generated IDs and hope for the best I don’t see how this can ever work robustly. Like symbol mangling, avoiding collisions requires using a function whose range has no overlap with the range of user-derived IDs. The only IDs I am aware of that are indescribable by the user are the empty string and strings with embedded null bytes. Neither of these seem like a workable idea. ---

**Author:** ![scnorth](https://sea2.discourse-cdn.com/graphviz/user_avatar/forum.graphviz.org/scnorth/32/89_2.png) [@scnorth](https://forum.graphviz.org/u/scnorth)\ **Post date:** [October 10, 2024, 3:12pm UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/13 "2024-10-10T15:12:08Z")

You’re right, of course, though also, nothing is absolutely robust. Like in C (and most languages), you can’t use a reserved word as a variable name, either. Yet somehow we carry on. ---

**Author:** ![smattr](https://sea2.discourse-cdn.com/graphviz/user_avatar/forum.graphviz.org/smattr/32/85_2.png) [@smattr](https://forum.graphviz.org/u/smattr)\ **Post date:** [October 11, 2024, 6:01am UTC](https://forum.graphviz.org/t/how-to-guarantee-unique-ids-in-svg-generated/2473/14 "2024-10-11T06:01:28Z")

> [@scnorth](#): > > nothing is absolutely robust Using the scheme I described above (incrementing `node%d`, but skipping any that collide with explicit IDs) where do we fail on robustness? > [@scnorth](#): > > Like in C (and most languages), you can’t use a reserved word as a variable name, either I don’t quite follow the analogy. In C, the standard clearly tells you which identifiers are reserved. In Graphviz, (AFAIK) there are no reserved IDs. At least [id | Graphviz](https://graphviz.org/docs/attrs/id/) does not document any.