Change default representation of edge attributes from vectors to lists & other miscellaneous adjustments #274

maxloeffler · 2024-12-05T09:48:36Z

Prerequisites

I adhere to the coding conventions (described here) in my code.
I have updated the copyright headers of the files I have modified.
I have written appropriate commit messages, i.e., I have recorded the goal, the need, the needed changes, and the location of my code modifications for each commit. This includes also, e.g., referencing to relevant issues.
I have put signed-off tags in all commits.
I have updated the changelog file NEWS.md appropriately.
I have checked whether I need to adjust the showcase file showcase.R with respect to my changes.
The pull request is opened against the branch dev.

Description

Replace deprecated igraph functions by their more recent counterparts.
Handle reading empty commit interaction data explicitly, to avoid crashing in that case.
To make splitting networks by ranges more consistent with splitting time-based, also add range information to the return value.
Convert certain edge attributes of networks to list type. This change is motivated by a more stricter type enforcement in igraph version 2.0 and up. When merging networks through igraph::disjoint_union, edge attributes that exist in both networks are now assumed to have identical types. But, when we simplify a network, its edge attributes convert to list type as they now need to store data that a vector in R cannot store anymore, i.e., one edge attribute now has to hold several lists of data of all the edges that got merged into one. Therefore, we need to convert edge attributes to list by default to be able to merge simplified and unsimplified networks.

Changelog

Added

Introduce convert.edge.attributes.to.list function.

Fixed

Fix reading empty commit interaction data.

Changed / Improved

Replace deprecated igraph functions.
Add range information to the return value in split.network.time.based.by.ranges, to be consistent with split.data.time.based.
Convert edge attributes to a list format to ensure compatibility with igraph.

codecov · 2024-12-05T10:10:45Z

Codecov Report

Attention: Patch coverage is 93.44262% with 4 lines in your changes missing coverage. Please review.

Project coverage is 81.09%. Comparing base (1d1fe7f) to head (3427e14).
Report is 1 commits behind head on dev.

Files with missing lines	Patch %	Lines
util-networks-metrics.R	0.00%	2 Missing ⚠️
util-networks.R	97.82%	1 Missing ⚠️
util-read.R	66.66%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##              dev     #274      +/-   ##
==========================================
+ Coverage   80.95%   81.09%   +0.14%     
==========================================
  Files          16       16              
  Lines        5051     5095      +44     
==========================================
+ Hits         4089     4132      +43     
- Misses        962      963       +1

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

util-split.R

Signed-off-by: Maximilian Löffler <[email protected]>

The interaction reading breaks if an interaction is empty in the data. This commit adds a check for empty interactions. Signed-off-by: Christian Hechtl <[email protected]>

Since version 2.0.0 of package `igraph`, the function `igraph::power_law_fit` (and its deprecated alias `igraph::fit.power.law`) does not automatically compute the p-value any more, which is needed by `metrics.scale.freeness`. In version 2.1.1 of package `igraph`, they have added a new parameter `p.value` to provide the possibility to enable p-value computation again. In addition, they have added a parameter `p.precision` which defaults to a precision of 0.01. To make `metrics.scale.freeness` work again with the most recent version of package `igraph`, the deprecated function calls to `igraph::fit.power.law` are replaced by `igraph::power_law_fit` and are adjusted to use the new parameter `p.value` to automatically compute the p-value again. However, we do not make use of `p.precision` and rely on the default precision instead. Signed-off-by: Thomas Bock <[email protected]>

util-networks.R

bockthom

While looking over the rest of this PR, I only found a couple of minor issues, please find the individual comments below.

Regarding the uncovered lines reported by codecov: I suggest to ignore them. 2 out of 4 missing lines are part of an untested module anyway. The other two lines refer to very special cases that have been previously untested too.

util-networks.R

util-split.R

tests/test-networks-equal-constructions.R

util-networks.R

tests/test-networks.R

This change makes splitting by ranges more consistent with time-based splitting. This works towards se-sic#273. Signed-off-by: Maximilian Löffler <[email protected]>

util-networks.R

Since igraph version 2.1, when joining networks using 'igraph::disjoint_union', edge attributes of the joining networks require identical types. As simplifiying networks necessarily converts types of edge attributes to list when merging edges, attributes now have to be of type list by default. Edge attributes that are explicitly considered during simplification and, therefore, are not converted to lists are excluded from this rule. This works towards fixing se-sic#271. Signed-off-by: Maximilian Löffler <[email protected]>

Adjust the tests in accordance to converting edge attributes to list type in the implementation. This works towards fixing se-sic#271. Signed-off-by: Maximilian Löffler <[email protected]>

This is necessary since igraph falsely fills in non-present edge attributes with NULLs instead of NAs in certain cases when using 'igraph::disjoint_union' and 'igraph::add_edges'. Signed-off-by: Maximilian Löffler <[email protected]>

'plyr::rbind.fill' uses NULL to fill missing values in lists. As we now use lists for most edge attributes, we need to handle this case separately to ensure missing values are filled with NAs instead. To fix this issue, we need to instantiate missing columns in dataframes with NAs before calling 'plyr::rbind.fill'. This operation is constant with respect to the amount of rows and should not impact performance too much. This works towards fixing se-sic#271. Signed-off-by: Maximilian Löffler <[email protected]>

This works towards fixing se-sic#271. Signed-off-by: Maximilian Löffler <[email protected]>

bockthom

I've a few comments regarding the news. Please find them below.

NEWS.md

bockthom · 2024-12-16T21:19:51Z

NEWS.md


 ### Fixed

 - Fix the creation of edgelists for issue-based artifact-networks by correctly iterating over the issue data (PR #264, 321d85043112971c04998249c14a0677a32c9004)
 - Fix a bug in `extract.timestamps` that occurs when the first `data.source` contains empty data and that leads to a return value of type numeric which should be POSIXct (PR #270, 10696e4cf4ae92371917ed8ccaec2b0183da145c, 646c01a42ad8decfbc9040030e790e51cb65cffd)
+- Fix `read.commit.interactions` by explicitly considering empty commit interaction data (PR #274, f591528a0f1f11b1a4390949ab770f3f74a766f9)


empty commit interaction data ➡️ non-existent commit interactions

(for explanation: the commit-interaction data per se are present, but specific interactions between specific commits are empty, that is, non-existent)

util-split.R

NEWS.md

Signed-off-by: Maximilian Löffler <[email protected]>

The igraph maintainers decided that they will not re-install consistency with their older versions (in terms of filling in non-existent values in 'igraph::disjoint_union' and 'igraph::add_edges' with NAs) yet (igraph/rigraph#1587). The previously temporary fix is therefore now semi-permanent. Signed-off-by: Maximilian Löffler <[email protected]>

Signed-off-by: Maximilian Löffler <[email protected]>

maxloeffler · 2025-01-12T16:00:13Z

Between the last and the second to last push I forgot to update the copyright headers of util-networks.R 😅.

bockthom

I just found a few minor inconsistencies in the README, please find them below.
If also @hechtlC is fine with the new announcement, this PR is ready to be merged then.

bockthom · 2025-01-12T19:58:02Z

NEWS.md

@@ -4,6 +4,10 @@

 ## unversioned

+### Announcement
+
+- Coronet is not compatible anymore with igraph versions below 2.1.0. This is due to the simultaneous deprecation of `subgraph.edges` and the introduction of the replacement for it, `subgraph_from_edges`, in igraph version 2.1.0.


Could you please use back ticks for the occurrences of igraph, as done in previous announcements?

bockthom · 2025-01-12T20:00:57Z

NEWS.md

@@ -18,8 +22,9 @@

 ### Changed/Improved

+- **Breaking Change**: Change the default representation of edge attributes from vectors to lists. This change is necessary for the interplay of coronet networks with certain igraph functionality since igraph version 2.1. (PR #274, 1c35d1fa2548deb297dbfa5e2b07fce31962c5b7, eda30b838369ec46376812298a3ea8159eec5789, 44c7b72e3234cb332bb2713fb408c124e67255d9, 7303eabef6a78198575fe5bdfc02813fde3d3974)


Same as above, please use back ticks for igraph also here. And my be we should also use version 2.1.0 here, to be consistent with the announcement.

bockthom · 2025-01-12T20:05:29Z

NEWS.md

@@ -28,6 +33,9 @@

 - Fix the creation of edgelists for issue-based artifact-networks by correctly iterating over the issue data (PR #264, 321d85043112971c04998249c14a0677a32c9004)
 - Fix a bug in `extract.timestamps` that occurs when the first `data.source` contains empty data and that leads to a return value of type numeric which should be POSIXct (PR #270, 10696e4cf4ae92371917ed8ccaec2b0183da145c, 646c01a42ad8decfbc9040030e790e51cb65cffd)
+- Fix `read.commit.interactions` by explicitly considering non-existent commit interaction data (PR #274, f591528a0f1f11b1a4390949ab770f3f74a766f9)
+- Add range information to network-splits when splitting a network using `split.network.time.based.by.ranges`. This effect also propagates into `split.networks.time.based` (PR #274, 87911ade231c44b93be194a1d6734f7de043a4af)
+- Adjust `metrics.scale.freeness` and `metrics.is.scale.free` functions be compatible with both older and newer igraph versions (PR #274, 4b0d5221dd56bb3c9ddf196f67719d4f503d9b61)


shouldn't it be "to be compatible with" instead of just "be compatible with" ?

bockthom · 2025-01-12T20:08:09Z

NEWS.md

@@ -28,6 +33,9 @@

 - Fix the creation of edgelists for issue-based artifact-networks by correctly iterating over the issue data (PR #264, 321d85043112971c04998249c14a0677a32c9004)
 - Fix a bug in `extract.timestamps` that occurs when the first `data.source` contains empty data and that leads to a return value of type numeric which should be POSIXct (PR #270, 10696e4cf4ae92371917ed8ccaec2b0183da145c, 646c01a42ad8decfbc9040030e790e51cb65cffd)
+- Fix `read.commit.interactions` by explicitly considering non-existent commit interaction data (PR #274, f591528a0f1f11b1a4390949ab770f3f74a766f9)


Here it seems to me that you have only fixed the first part (replaced empty be non-existent), but this only makes sense if you also replace the following words as indicated in my previous review: commit interaction data ➡️ commit interactions.

(the commit interaction data are present, but no commit interactions are present).

bockthom reviewed Dec 5, 2024

View reviewed changes

util-split.R Outdated Show resolved Hide resolved

maxloeffler and others added 3 commits December 7, 2024 17:06

Replace deprecated igraph functions

e3617b8

Signed-off-by: Maximilian Löffler <[email protected]>

Fix read.commit.interactions()

f591528

The interaction reading breaks if an interaction is empty in the data. This commit adds a check for empty interactions. Signed-off-by: Christian Hechtl <[email protected]>

maxloeffler force-pushed the dev branch from e3050e1 to 0743294 Compare December 7, 2024 16:07

bockthom reviewed Dec 9, 2024

View reviewed changes

util-networks.R Outdated Show resolved Hide resolved

bockthom reviewed Dec 9, 2024

View reviewed changes

Add range information to splits in 'split.network.time.based.by.ranges'

87911ad

This change makes splitting by ranges more consistent with time-based splitting. This works towards se-sic#273. Signed-off-by: Maximilian Löffler <[email protected]>

bockthom reviewed Dec 10, 2024

View reviewed changes

util-networks.R Show resolved Hide resolved

maxloeffler added 5 commits December 10, 2024 11:21

Adjust tests to expect edge attributes as lists

eda30b8

Adjust the tests in accordance to converting edge attributes to list type in the implementation. This works towards fixing se-sic#271. Signed-off-by: Maximilian Löffler <[email protected]>

Add test for 'convert.edge.attributes.to.list'

7303eab

This works towards fixing se-sic#271. Signed-off-by: Maximilian Löffler <[email protected]>

maxloeffler force-pushed the dev branch from 0743294 to 7303eab Compare December 10, 2024 10:21

maxloeffler changed the title ~~Miscellaneous fixes and adjustments~~ Change default representation of edge attributes from vectors to lists & other miscellaneous adjustments Dec 16, 2024

bockthom requested changes Dec 16, 2024

View reviewed changes

bockthom added this to the v5.0 milestone Dec 17, 2024

Update 'NEWS.md'

2d2a5f6

Signed-off-by: Maximilian Löffler <[email protected]>

maxloeffler force-pushed the dev branch from 260af89 to 440c701 Compare January 12, 2025 15:55

maxloeffler added 3 commits January 12, 2025 16:58

Bump required version of 'igraph' in 'README.md' to 2.1.0

6c3bcd1

Signed-off-by: Maximilian Löffler <[email protected]>

Update 'NEWS.md'

3427e14

Signed-off-by: Maximilian Löffler <[email protected]>

maxloeffler force-pushed the dev branch from 440c701 to 3427e14 Compare January 12, 2025 15:59

bockthom reviewed Jan 12, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Change default representation of edge attributes from vectors to lists & other miscellaneous adjustments #274

Change default representation of edge attributes from vectors to lists & other miscellaneous adjustments #274

maxloeffler commented Dec 5, 2024 •

edited

Loading

codecov bot commented Dec 5, 2024 •

edited

Loading

bockthom left a comment

bockthom left a comment

bockthom Dec 16, 2024

maxloeffler commented Jan 12, 2025

bockthom left a comment

bockthom Jan 12, 2025

bockthom Jan 12, 2025

bockthom Jan 12, 2025

bockthom Jan 12, 2025

		@@ -18,8 +22,9 @@

		### Changed/Improved

		- Breaking Change: Change the default representation of edge attributes from vectors to lists. This change is necessary for the interplay of coronet networks with certain igraph functionality since igraph version 2.1. (PR #274, 1c35d1fa2548deb297dbfa5e2b07fce31962c5b7, eda30b838369ec46376812298a3ea8159eec5789, 44c7b72e3234cb332bb2713fb408c124e67255d9, 7303eabef6a78198575fe5bdfc02813fde3d3974)

Change default representation of edge attributes from vectors to lists & other miscellaneous adjustments #274

Are you sure you want to change the base?

Change default representation of edge attributes from vectors to lists & other miscellaneous adjustments #274

Conversation

maxloeffler commented Dec 5, 2024 • edited Loading

Prerequisites

Description

Changelog

Added

Fixed

Changed / Improved

codecov bot commented Dec 5, 2024 • edited Loading

Codecov Report

bockthom left a comment

Choose a reason for hiding this comment

bockthom left a comment

Choose a reason for hiding this comment

bockthom Dec 16, 2024

Choose a reason for hiding this comment

maxloeffler commented Jan 12, 2025

bockthom left a comment

Choose a reason for hiding this comment

bockthom Jan 12, 2025

Choose a reason for hiding this comment

bockthom Jan 12, 2025

Choose a reason for hiding this comment

bockthom Jan 12, 2025

Choose a reason for hiding this comment

bockthom Jan 12, 2025

Choose a reason for hiding this comment

maxloeffler commented Dec 5, 2024 •

edited

Loading

codecov bot commented Dec 5, 2024 •

edited

Loading