Docs/adr-lpmstorage by VikiPeeva · Pull Request #9 · promworkbench/LocalProcessModelDiscoveryByCombiningPlaces

VikiPeeva · 2026-06-20T14:25:32Z

No description provided.

Added reviewers section to the ADR document.

alkuzman

I did not review the proposals (options) in detail. I will do that in another round of review.

alkuzman · 2026-06-22T03:40:15Z

+Superseded by:
+---
+
+# ADR-0001: Data Storage Format


The title is imprecise. Storage format of what exactly?

Also, the numbers of the ADRs should be stable over time. If you have PRs for multiple ADRs you don't know which one you will merge first therefore assigning numbers becomes very hard. The best way could be to assign the number of the PR that creates the ADR. For example now you created a draft, now you can update both the file and the titple to ADR-0009. That way you will always be sure that they have unique numbers.

I don't see the benefit of this. Why wouldn't they be stable with the other strategy, numbering them in order of creation? The way you propose it, there will be gaps in the numbers, since not all pull requests are for an ADR.

The title has been updated.

alkuzman · 2026-06-22T03:43:23Z

+
+## Context
+
+We need to store _Local Process Models (LPMs) and their occurrence lists in an event log_. Event log sizes can vary, 


Maybe you should answer the following questions first:

Why not jus LPMs?

why occurrences?

why in an event log?

alkuzman · 2026-06-22T03:49:47Z

+- Query-based read/write
+
+There also exist various ways one can align LPMs to an event log:
+- The storage focuses on one variation


You use the word variation here without explaining what is it. The image is not enough to explain. Also, why having this constraint is important?

Example variations are now listed to clarify the problem. The comment regarding why this constraint is important is unclear to me.

alkuzman

In general make it clear what files would be generated for each option. Maybe provide a example tree structures for each.

in Decision Drivers give list of usecases that will require querying or writing from this format. Maybe shortly discuss from what option they might benefit.

alkuzman · 2026-07-05T12:42:07Z

+Superseded by:
+---
+
+# ADR-0001: Data Storage Format


Also, the numbers of the ADRs should be stable over time. If you have PRs for multiple ADRs you don't know which one you will merge first therefore assigning numbers becomes very hard. The best way could be to assign the number of the PR that creates the ADR. For example now you created a draft, now you can update both the file and the titple to ADR-0009. That way you will always be sure that they have unique numbers.

alkuzman · 2026-07-05T12:56:58Z

+     "t1": {
+       "e1": "t_a", 
+       "e2": ">>",
+       "e3": "t_a",


How could $t_a$ happen twice?

Once it happens because of one occurrence list and the other because of another. The example above is now more detailed, and that is what I meant in the comment above: the alignment in the file actually serves as a collective alignment for all occurrence lists for that trace and local process model.

alkuzman · 2026-07-05T12:57:45Z

+       "e5": "t_d"
+     }, 
+     "t2": {
+       "e1": ">>",


why $e_1$ is not matched with $t_a$?

It is not matched because there is no alignment mapping an entire model execution to the trace, since $b$ is missing from trace $t2$.

alkuzman · 2026-07-05T13:06:13Z

+- Using only existing formats
+
+**Cons**
+- Complete occurrence lists need to be reconstructed


Is this really expensive thing to do? To me it seems like a straightforward operation that can be done on demand.
However, this might become too expensive if the log is too big and you need to filter by few LPMs. Then Option C becomes really attractive.

An explanation regarding why this is expensive has now been added.

alkuzman · 2026-07-05T13:07:08Z

+### Option B: Separate Full Alignments File in Human-Readable Format (e.g., JSON)
+
+For each set of models for which an occurrence list is computed, a report is generated that includes:
+- lpm set info,


I don't understand what this is.

In the update, some examples of what can be included here are given.

alkuzman · 2026-07-05T13:08:01Z

+
+$lpm2$: a -> b -> d
+
+```json


It is not clear that there could be multiple such files.

I don't understand what such files refer to.

alkuzman · 2026-07-05T13:13:47Z

+To be able to store additional attributes, a new format for the models can be introduced or additional json file 
+can be used only for storing the additional attributes.
+
+### Option B: Separate Full Alignments File in Human-Readable Format (e.g., JSON)


This cannot handle multiple alignments for the same trace (events) therefore it should not be considered.

Actually, it can handle multiple alignments; it is like a global alignment for all possible alignments for that trace. If an event can be aligned as part of any of the occurrences, it will be shown as a synchronous move. One problem is that the occurrence lists need to be reconstructed similarly to Option A. The second problem is that if I want to align the same event to different transitions for different occurrences, this is not possible.

VikiPeeva and others added 4 commits April 17, 2026 16:54

initial version of the adr for lpm-log-alignment storage

7831733

Remove reviewers from ADR-0001

ba6bba1

Update ADR-0001 to have basic info in frontmatter

285e536

Added reviewers section to the ADR document.

Reformat ADR template to contain frontmatter

bb1ed5f

VikiPeeva requested a review from alkuzman June 20, 2026 14:25

VikiPeeva self-assigned this Jun 20, 2026

VikiPeeva added the documentation Improvements or additions to documentation label Jun 20, 2026

alkuzman suggested changes Jun 22, 2026

View reviewed changes

alkuzman suggested changes Jul 5, 2026

View reviewed changes

VikiPeeva added 3 commits July 7, 2026 11:03

some updates considering the feedback in #9 and clarification added

6924839

small structuring fix

e23f6f5

Added sections as suggested in #9

405235b


		## Context

		We need to store _Local Process Models (LPMs) and their occurrence lists in an event log_. Event log sizes can vary,


		$lpm2$: a -> b -> d

		```json

Uh oh!

Conversation

VikiPeeva commented Jun 20, 2026

Uh oh!

alkuzman left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

VikiPeeva Jul 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

alkuzman Jun 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

alkuzman left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

VikiPeeva Jul 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

VikiPeeva Jul 7, 2026 •

edited

Loading

alkuzman Jun 22, 2026 •

edited

Loading

VikiPeeva Jul 7, 2026 •

edited

Loading