feat: New `ns_graph()` and lower-level `ns_df()` and `ns_metadata()` to download from [netzschleuder](https://networks.skewed.de) #23

schochastics · 2025-04-05T18:29:11Z

take care of https://networks.skewed.de/restricted
add custom rate limits (current implementation doesnt seem to work)

Run: https://github.com/igraph/igraphdata/actions/runs/14284649633

szhorvat · 2025-04-05T18:38:56Z

Is it documented how networks are encoded into CSV on Netzschleuder, and if not, can we write a short description and include it in comments? I expect this won't be as simple to get working with the majority of dataset as it might seem.

R/netzschleuder.R

szhorvat · 2025-04-05T18:43:14Z

Here's a stress-test ;-) Not too large, but plenty of attributes. https://networks.skewed.de/net/blumenau_drug

schochastics · 2025-04-05T18:56:24Z

Is it documented how networks are encoded into CSV on Netzschleuder, and if not, can we write a short description and include it in comments? I expect this won't be as simple to get working with the majority of dataset as it might seem.

My plan is to write a script that runs through all networks once to see if there are any differences. My sampling so far showed that it looks pretty consistent

edges.csv: source,target, attributes
nodes.csv: name, attributes
gprobs.csv: same set of metadata

szhorvat · 2025-04-06T20:08:47Z

Here's a stress-test ;-) Not too large, but plenty of attributes. https://networks.skewed.de/net/blumenau_drug

This one has a logical vertex attribute that is not currently imported as such. I'm not sure how easy it is to detect that this is logical while maintaining good performance.

schochastics · 2025-04-07T06:16:04Z

Here's a stress-test ;-) Not too large, but plenty of attributes. networks.skewed.de/net/blumenau_drug

This one has a logical vertex attribute that is not currently imported as such. I'm not sure how easy it is to detect that this is logical while maintaining good performance.

Good catch. I would add minty as a dependency for this

maelle

🎉 I find that the function name is a mouthful 😸

R/netzschleuder.R

Co-authored-by: Maëlle Salmon <[email protected]>

Run: https://github.com/igraph/igraphdata/actions/runs/14379148488

schochastics · 2025-04-10T11:30:40Z

@maelle how do I properly document the dots argument? check complains about them

schochastics · 2025-04-10T12:19:13Z

The function now has custom ratelimiting. 20 calls per minute. Is that reasonable @szhorvat?

schochastics · 2025-04-10T18:05:37Z

🎉 I find that the function name is a mouthful 😸

Indeed. Do you have a suggestion? I don't know how to make it shorter. graph_schleuder?

szhorvat · 2025-04-10T18:37:35Z

Some comments from my side:

Name: I think the name is good. It should be descriptive since it will be a relatively rarely seen function. Auto-complete makes this a non-issue.

Or do you mean that Netzschleuder is just hard to spell for people with no exposure to German?

directed parameter: I suppose you added this because the CSV format does not encode directedness? This must not be left to the user. The API allows for retrieving directedness, see e.g. https://networks.skewed.de/api/net/advogato The value from there can be used.

I get a warning with some networks, e.g. graph_from_netzschleuder('packet_delays')

Warning message:
`type_convert()` only converts columns of type 'character'.
- `df` has no columns of type 'character'

Multi-network datasets:

These do not work currently, this is expected, right?

When they do start working, what will happen if only the collection name is given, not a subnetwork?

Can we support the slash syntax, like here? graph_from_netzscheluder('fresh_webs/AkatoreA')

Is it very un-R-like to also support the tuple syntax, graph_from_netzscheluder(c('fresh_webs', 'AkatoreA'))? This is friendly for programmatic access. It would be idiomatic both in Python and Mathematica.

With these approaches we could get rid of the network parameter completely and have a single descriptor for any given network. I don't like the network parameter because some data can be downloaded by specifying a single parameter, some by specifying two. It's better to have a single identifier that will always work, for any network.

schochastics · 2025-04-10T19:16:59Z

Thanks for the feedback!

directed parameter

Yes that was due to the csv format and me not wanting to add dependency when this was still planned as an igraph function. I will implement this via the API. Same as bipartite

warning with some networks

Thanks, thats fixable.

Multi-network datasets

Do you mean that it should be possible to download a whole collection at once? Like a list of all fresh_webs networks? If we allow full collection download, we have to block some, like openstreetmaps. Thats just too huge. Maybe this goes in a separate function?

You are right currently we need collection name+network name to get one network from a collection. For consistency with gt, I think I will go with the slash notations.

Is it very un-R-like to also support the tuple syntax

Yes. When I saw this, I immediately thought you want to fetch two networks, not fresh_webs/AkatoreA

szhorvat · 2025-04-10T19:27:01Z

Do you mean that it should be possible to download a whole collection at once?

No, I didn't meant to imply that, I was just asking about what your plans were. I vote for not allowing a full set download. But we still need to choose some response to providing only a set name instead of a single network.

This could be a simple not found, but perhaps it's nicer to tell the user to choose a specific network? Maybe the error message could even link to the relevant networks.skewed.de dataset page, to aid the user in selecting one?

schochastics · 2025-04-10T19:31:42Z

ok yes I agree that full collection shouldnt be possible. I will think of a logic for this

szhorvat · 2025-04-10T19:42:50Z

Some more brainstorming: Now we have read_from_netzschleuder, which gives a dataframe, and graph_from_netzschleuder, which gives a network.

"read" to me does not say "dataframe", but perhaps this is because of my inexperience with R. There are "read" functions that don't produce dataframes, such as read_graph in igraph.

If we could find a more descriptive name that suggests "this downloads network data", then we can shorten netzschleuder to ns (for people who are scared of consonant clusters 😆 ), as it will no longer be how people recognize the function. There will also be a possibility for future additions of other data sources. If R were used by mathematicians, I'd say house of graphs :-)

The output format doesn't necessarily need to be split between two functions. It can also be controlled by a parameter, in which case we could have a single function. For example, network_from_ns('foo', output='graph') could get the graph, output='data.frame' would get a dataframe and output='metadata' would get the metadata returned by the API (no graph).

network_from is not a pretty name and I'm not totally happy with it, but it has the advantages:

it's not confused with a typical graph_from igraph function; this is an igraphdata function, which behaves differently, and can output multiple things in multiple formats
it's descriptive enough that netzschleuder could be shortened to ns without confusion

Other prefix ideas instead of network_from:

read_graph_from
get_graph_from
get_network
net_from
...

schochastics · 2025-04-10T20:20:08Z

While you brainstormed, I implemented what we discussed before. Now we have

pkgload::load_all("~/git/R_packages/igraphdata/")
#> ℹ Loading igraphdata
graph_from_netzschleuder("blumenau_drug")
#> IGRAPH 2849f38 UNW- 75 181 -- 
#> + attr: name (v/c), p_node (v/n), hormone (v/l), len_i (v/n), color
#> | (v/c), label (v/c), class (v/c), x (v/n), y (v/n), tau (e/n),
#> | severity (e/c), weight (e/n), color (e/c), gender (e/c),
#> | patients_norm (e/n), RRI.F (e/n), patients (e/n), tau_norm (e/n),
#> | RRI.M (e/n)
#> + edges from 2849f38 (vertex names):
#>  [1] DB01174--DB00741 DB01174--DB00860 DB01174--DB00635 DB01174--DB00571
#>  [5] DB01174--DB00367 DB01174--DB01223 DB00741--DB00252 DB00252--DB00537
#>  [9] DB00252--DB01223 DB00252--DB00695 DB00252--DB00359 DB00252--DB00338
#> [13] DB00252--DB00196 DB00252--DB01234 DB00252--DB00860 DB00252--DB00829
#> + ... omitted several edges
graph_from_netzschleuder("fresh_webs/AkatoreA")
#> IGRAPH 58361eb DNW- 85 227 -- 
#> + attr: name (v/c), x (v/n), y (v/n), weight (e/n)
#> + edges from 58361eb (vertex names):
#>  [1] 1 ->45 1 ->46 1 ->47 1 ->48 1 ->49 1 ->51 1 ->52 1 ->53 1 ->54 1 ->55
#> [11] 1 ->56 1 ->57 1 ->58 1 ->63 1 ->64 1 ->66 1 ->67 1 ->68 1 ->69 1 ->70
#> [21] 1 ->71 1 ->72 1 ->73 1 ->74 1 ->75 1 ->76 1 ->78 1 ->81 1 ->83 1 ->84
#> [31] 1 ->85 2 ->83 3 ->46 3 ->47 3 ->48 3 ->50 3 ->53 3 ->54 3 ->55 3 ->56
#> [41] 3 ->58 3 ->59 3 ->60 3 ->61 3 ->62 3 ->65 3 ->67 3 ->68 3 ->70 3 ->71
#> [51] 3 ->72 3 ->74 3 ->75 3 ->78 3 ->82 3 ->83 4 ->45 4 ->46 4 ->66 4 ->80
#> [61] 5 ->83 6 ->46 6 ->48 6 ->50 6 ->53 6 ->55 6 ->61 6 ->65 6 ->67 6 ->69
#> [71] 6 ->75 6 ->83 7 ->55 7 ->61 7 ->83 8 ->55 8 ->83 9 ->83 10->71 10->72
#> + ... omitted several edges
graph_from_netzschleuder("fresh_webs")
#> Error in `read_from_netzschleuder()` at igraphdata/R/netzschleuder.R:139:3:
#> ! fresh_webs is a collection and downloading a whole collection is not
#>   permitted.
#> ℹ see <https://networks.skewed.de/net/fresh_webs>
graph_from_netzschleuder("fresh_webs/nope")
#> Error in `read_from_netzschleuder()` at igraphdata/R/netzschleuder.R:139:3:
#> ! nope is not part of the collection fresh_webs.
#> ℹ see <https://networks.skewed.de/net/fresh_webs>

^{Created on 2025-04-10 with reprex v2.1.1}

schochastics · 2025-04-10T20:23:40Z

This is maybe personal, but I do not like it if the output of a function changes depending on a parameter like network_from_ns('foo', output='graph') would do.

I am not good in naming things, so I am not sure what the best option would be.

Run: https://github.com/igraph/igraphdata/actions/runs/14495485835

DESCRIPTION

krlmlr

Thanks, a very useful function. Some general remarks:

Do we need to close the connections opened with unz() to avoid warnings? An easy way is withr::local_connection(unz(...)) if we don't mind the dependency.
I prefer [[ over [ for accessing single elements of a vector, and vapply() over sapply() : Strictness
Should we use the longer netzschleuder_ prefix instead of ns_ ?
I wonder if the df and graph functions should accept a metadata object. We'd add a class and provide a print() method for extra sugar.

krlmlr · 2025-04-16T19:41:56Z

R/netzschleuder.R

+  if (net_ident[1] == net_ident[2] && length(unlist(raw$nets)) > 1) {
+    cli::cli_abort(
+      c(
+        "{net_ident[1]} is a collection and downloading a whole collection is not permitted.",
+        "i" = "see {.url {collection_url}}"
+      )
+    )
+  } else if (net_ident[1] == net_ident[2]) {
+    return(raw)
+  } else {
+    idx <- which(unlist(raw[["nets"]]) == net_ident[2])
+    if (length(idx) == 0) {
+      cli::cli_abort(
+        c(
+          "{net_ident[2]} is not part of the collection {net_ident[1]}.",
+          "i" = "see {.url {collection_url}}"
+        ),
+        call = call
+      )
+    }
+    raw[["analyses"]] <- raw[["analyses"]][[net_ident[2]]]
+    raw[["nets"]] <- raw[["nets"]][idx]
+    raw
+  }


Do you have examples for each of the code paths?

Yes all three are needed and The error handling should tell users what they are supposed to do to get the correct network

R/netzschleuder.R

krlmlr · 2025-04-16T19:46:03Z

R/netzschleuder.R

+  if (grepl("/", x)) {
+    return(strsplit(x, "/", fixed = TRUE)[[1]])
+  } else {
+    c(x, x)


This looks interesting to me. I wonder if we can make the intent here clearer.

This mechanism is need for the API. If a collection has only one network, the url is
https://networks.skewed.de/net/7th_graders/files/7th_graders.csv.zip. SO the name is duplicated. For a network in a collection, the zip file has a different name

Got it. Should we then return a named list instead, to clarify the intent?

sounds good

krlmlr · 2025-04-16T19:46:51Z

R/netzschleuder.R

+
+download_file <- function(zip_url, token = NULL, file) {
+  resp <- make_request(zip_url, token)
+  writeBin(httr2::resp_body_raw(resp), file)


Are the files always small enough to fit into memory?

There are some networks that go up to 10GB. Should we implement something that stops users to download files over a certain size?

Should we implement something that stops users to download files over a certain size?

IMO it's not a bad idea. Many of our users are inexperienced, and the risk of accidentally trying to download something large is too high. But is it possible the determine the size in advance? Or would you use the edge and vertex count as a proxy for the size?

Do I understand correctly that httr2 does automatically cache requests, preventing the re-downloading of large data?

If you do this, I'd suggest having an option to override and download anyway.

Something to consider is whether R has limits on the number of elements in dataframes, or number of rows it can read from CSVs. R still has this silly limitation of 32-bit integers, and while it can index with doubles now (weird!), not every object has support for such indexing. E.g. matrices don't.

This is about raw data loaded into memory, which is what it looks like this code is doing. Can we let curl handle the downloading in this case, to write directly to disk in chunks? There might be a way to show a progress bar too, and the user can cancel if it's too large.

I will dig into httr2 a bit on this. I would like to keep using it here too for the rate limiting. I am sure there is something that can help us here. Otherwise my idea was to create an option or so that specifies a hard upper limit of edges/nodes that are permitted.

Is the rate limiting important for downloads, or just for API calls?

We can probably get the body in chunks and write piecemeal if really necessary.

The rate limiting is the most important for files because that is the heaviest load on the server.

DESCRIPTION

R/zzz.R

schochastics · 2025-04-17T08:01:32Z

withr::local_connection(unz(...))

shouldn't on.exit(unlink(temp)) take care of that?

Should we use the longer netzschleuder_ prefix instead of ns_ ?

We discussed it further up that ns is prefered.

I wonder if the df and graph functions should accept a metadata object. We'd add a class and provide a print() method for extra sugar.

Good idea. I will explore that

Run: https://github.com/igraph/igraphdata/actions/runs/14510993378

schochastics · 2025-04-17T09:06:22Z

Not sure the switch to a S3 method for ns_graph is good. The code gets a bit convoluted since all error handling is done by ns_meta to minimize API calls. ns_meta is also never called in ns_graph but indirectly via ns_df. So in my opinion, the current way is the cleanest, although I get that a S3 method and a "meta print" function would be nice

szhorvat · 2025-04-17T09:52:46Z

I worked a bit more on trying to break things, and here's what I found:

Note that ns_graph('ambassador/KINSHIP') is correct syntax and downloads the network.

Let's misspell the subnetwork:

> ns_graph('ambassador/KINSIP')
Error in `rlang::abort()`:
! `call` must be a call or environment, not a primitive
  function.
Run `rlang::last_trace()` to see where the error occurred.

This error could be nicer.

Let's write nothing after /:

> ns_graph('ambassador/')
Error in if (net_ident[1] == net_ident[2] && length(unlist(raw$nets)) >  : 
  missing value where TRUE/FALSE needed

Let's add one more slash and an extra name: ns_graph('ambassador/KINSHIP/foo'). This behaves the same as ns_graph('ambassador/KINSHIP'). It should instead error.

Let's take a dataset that has only a single network, so it should not use the / syntax:

> ns_graph('packet_delays/foo')
Error in `rlang::abort()`:
! `call` must be a call or environment, not a primitive
  function.
Run `rlang::last_trace()` to see where the error occurred.

szhorvat · 2025-04-17T10:03:43Z

Should ns_metadata() work directly on collections instead of individual networks? For example, both ns_metadata('ambassador') and ns_metadata('ambassador/KINSHIP') could work. It seems to me that this might be useful.

schochastics · 2025-04-17T13:27:03Z

Thanks for the thorough testing. Now try to break it again ;-).

I think you are right with ns_metadata. I will implement that too

szhorvat · 2025-04-17T15:59:42Z

Now try to break it again ;-).

> ns_graph(c('a','b'))
Error in if (grepl("/", x)) { : the condition has length > 1

This sometimes happens when passing the wrong type, as that type of converted to character, which tends to result in a vector of more than one elements.

"foo/" is interpreted as "foo", but I'm not sure I would call this a real problem.

I'm out of ideas on how to break it.

Out of curiosity, are downloaded networks cached? It's my impression that httr2 does not automatically?

schochastics · 2025-04-17T17:03:45Z

I was thinking about caching. If I find a way that is not too convoluted, then I implement it. Httr2 has a cache function but not sure how persistent it is. Creating a persistent rate limit was already quite convoluted

szhorvat · 2025-04-17T17:58:23Z

This suggests that req_perform(), which you used, already does some caching. Perhaps that's sufficient.

https://httr2.r-lib.org/reference/req_cache.html

krlmlr

Thanks! A wide range from questions to nitpicks to more serious concerns.

krlmlr · 2025-04-17T18:39:04Z

DESCRIPTION

+    cli,
+    httr2,


All of those must be in "Suggests".

krlmlr · 2025-04-17T18:40:24Z

R/netzschleuder.R

+  #remove trailing /
+  x <- sub("/$", "", x)
+  #remove double slash
+  x <- sub("//", "/", x)


Why is this important?

because @szhorvat tried to break the name parameter and these are just some cleaning of common user input errors that could occur

krlmlr · 2025-04-17T18:40:40Z

R/netzschleuder.R

+  if (grepl("/", x)) {
+    return(strsplit(x, "/", fixed = TRUE)[[1]])
+  } else {
+    c(x, x)


Got it. Should we then return a named list instead, to clarify the intent?

krlmlr · 2025-04-17T18:43:01Z

R/netzschleuder.R

+
+download_file <- function(zip_url, token = NULL, file) {
+  resp <- make_request(zip_url, token)
+  writeBin(httr2::resp_body_raw(resp), file)


This is about raw data loaded into memory, which is what it looks like this code is doing. Can we let curl handle the downloading in this case, to write directly to disk in chunks? There might be a way to show a progress bar too, and the user can cancel if it's too large.

R/netzschleuder.R

krlmlr · 2025-04-17T18:55:06Z

R/netzschleuder.R

+  nodes_df_raw <- utils::read.csv(unz(temp, node_file_name))
+  #suppress warning if no character columns found
+  nodes_df <- suppressWarnings(minty::type_convert(nodes_df_raw))
+  names(nodes_df)[1] <- "id"


I'm leaning towards implementing lintr just for this, here and elsewhere.

Suggested change

names(nodes_df)[1] <- "id"

names(nodes_df)[[1]] <- "id"

R/zzz.R

krlmlr · 2025-04-17T19:03:29Z

R/netzschleuder.R

+#' @rdname netzschleuder
+#' @export
+ns_graph <- function(name, token = NULL) {
+  graph_data <- ns_df(name, token = token)


My comment was not about S3, we can implement poor man's dispatch here (hand-rolled check if it's a character or a classed object). Users may want to inspect or do other things to the meta object before creating a graph or a data frame from it. We could also decouple downloading from df creation, but that's a bit of a can of worms.

R/netzschleuder.R

Run: https://github.com/igraph/igraphdata/actions/runs/14625553900

schochastics · 2025-04-23T19:15:06Z

I think I have caught up with all suggestions now. The big ones:

a parameter that prevents downloading of too big files
metadata can be used as input to ns_graph() and has its own print function

krlmlr · 2025-04-28T19:48:44Z

Thanks! Opening a new issue to address remaining comments.

schochastics and others added 2 commits April 5, 2025 20:28

added support for netzschleuder

25f0193

chore: Auto-update from GitHub Actions

67ec488

Run: https://github.com/igraph/igraphdata/actions/runs/14284649633

szhorvat reviewed Apr 5, 2025

View reviewed changes

R/netzschleuder.R Outdated Show resolved Hide resolved

schochastics added 2 commits April 5, 2025 20:56

better source/target handling

5c5895d

added token for restricted networks

233b9e0

added minty for type checking

e521a71

maelle reviewed Apr 9, 2025

View reviewed changes

schochastics and others added 2 commits April 10, 2025 13:19

Apply suggestions from code review

4b07fc2

Co-authored-by: Maëlle Salmon <[email protected]>

chore: Auto-update from GitHub Actions

33ae5ed

Run: https://github.com/igraph/igraphdata/actions/runs/14379148488

schochastics added 3 commits April 10, 2025 13:32

changes of review

db8b0db

added missing quotes

51046b9

added proper httr2 requests with custom ratelimits

d8b1f77

schochastics added 3 commits April 10, 2025 22:12

added slash notation and directed/bipartite checking via metadata

bc6a34a

suppress minty warnings

d7a0b06

docs update

ca89264

schochastics and others added 3 commits April 16, 2025 16:38

added more comments for clarity

4431e4b

added David as ctb and a sentence about netzschleuder

a52dec8

chore: Auto-update from GitHub Actions

10b25c4

Run: https://github.com/igraph/igraphdata/actions/runs/14495485835

szhorvat approved these changes Apr 16, 2025

View reviewed changes

DESCRIPTION Outdated Show resolved Hide resolved

DESCRIPTION Show resolved Hide resolved

DESCRIPTION Show resolved Hide resolved

DESCRIPTION Outdated Show resolved Hide resolved

krlmlr reviewed Apr 16, 2025

View reviewed changes

schochastics and others added 3 commits April 17, 2025 10:03

review changes

4485280

chore: Auto-update from GitHub Actions

02fb9db

Run: https://github.com/igraph/igraphdata/actions/runs/14510993378

moved minty check to ns_df

0c717fc

better name error handling

1974d6a

krlmlr reviewed Apr 17, 2025

View reviewed changes

schochastics and others added 8 commits April 18, 2025 21:05

moved all pkgs to suggest and removed onLoad

29412cf

earlier renaming and better regex

0ad558d

double parenthesis

3cd0496

allow metadata as input to ns_graph

eb12a3a

chore: Auto-update from GitHub Actions

c9ef882

Run: https://github.com/igraph/igraphdata/actions/runs/14625553900

implement file size limit

94bab69

docs

18f87d1

added meta data print

2a6d9f1

krlmlr changed the title ~~feat: add support for netzschleuder~~ feat: New ns_graph() and lower-level ns_df() and ns_metadata() to download from [netzschleuder](https://networks.skewed.de) Apr 28, 2025

krlmlr merged commit d80f4b6 into main Apr 28, 2025
6 checks passed

feat: New ns_graph() and lower-level ns_df() and ns_metadata() to download from [netzschleuder](https://networks.skewed.de) #23

feat: New ns_graph() and lower-level ns_df() and ns_metadata() to download from [netzschleuder](https://networks.skewed.de) #23

Uh oh!

Conversation

schochastics commented Apr 5, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

szhorvat commented Apr 5, 2025

Uh oh!

Uh oh!

szhorvat commented Apr 5, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

schochastics commented Apr 5, 2025

Uh oh!

szhorvat commented Apr 6, 2025

Uh oh!

schochastics commented Apr 7, 2025

Uh oh!

maelle left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

schochastics commented Apr 10, 2025

Uh oh!

schochastics commented Apr 10, 2025

Uh oh!

schochastics commented Apr 10, 2025

Uh oh!

szhorvat commented Apr 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

schochastics commented Apr 10, 2025

Uh oh!

szhorvat commented Apr 10, 2025

Uh oh!

schochastics commented Apr 10, 2025

Uh oh!

szhorvat commented Apr 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

schochastics commented Apr 10, 2025

Uh oh!

schochastics commented Apr 10, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

krlmlr 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!

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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

feat: New `ns_graph()` and lower-level `ns_df()` and `ns_metadata()` to download from [netzschleuder](https://networks.skewed.de) #23

feat: New `ns_graph()` and lower-level `ns_df()` and `ns_metadata()` to download from [netzschleuder](https://networks.skewed.de) #23

schochastics commented Apr 5, 2025 •

edited

Loading

szhorvat commented Apr 5, 2025 •

edited

Loading

szhorvat commented Apr 10, 2025 •

edited

Loading

szhorvat commented Apr 10, 2025 •

edited

Loading

schochastics commented Apr 17, 2025 •

edited

Loading

szhorvat commented Apr 17, 2025 •

edited

Loading