gt_add_meta Added to add custom meta data to nodes.gt_preproc_edges Added to preprocess edges before comoputing nodes.gt_co_edges replaces gt_edges_hashes, it now works with other columns than just hashtags.gt_edges_hash is deprecated in favour of gt_edges.gt_edges_bind and gt_co_edges_bind added to bind edges together and build more complex graphs._ are no longer necessary and are thus deprecated.The origin gt_edges functions, first part of the package in 2014, was extracting @tagged users from tweets’ text with convoluted regular expressions: it is no longer necessary as rtweet now returns mentions_screen_name, hence the aforementioned changes.
gt_edges_hashes_ and gt_edges_hashes to build networks of #hashtags co-mentions.%<-% from the zeallot package to unpack the nodes and edges.gt_dyn bug where lifetime was not working properly.Removed splitstackshape dependency ahead of its archival; now uses tidyr.
gt_nodes returns number of n_edges, the number of edges the node is present in.gt_edges_hash and respective escape hatch addedgt_dyn returns correct start and end.Major release: overhaul to 1) make computations much faster, 2) make the whole package more tidyverse friendly and 3) switch to rtweet as main source.
getEdges & getNodes are now deprecated in favour of gt_edges and gt_nodes
dynamise deprecated in favour of gt_dyn
magrittr pipe added.gt_collect added: use to get to collect edges and nodes as list.gt_graph added: use to convert to igraph object.Performance
library(graphTweets)
library(rtweet)
token <- create_token("APP", "xxxXXxxxx", "xXXXxxXX")
tweets <- search_tweets("#rstats", token = token)
rbenchmark::benchmark(
"v3.2" = {
edges <- getEdges(as.data.frame(tweets), "screen_name", "text")
g <- igraph::graph.data.frame(edges, TRUE)
},
"v4" = {
tweets %>%
gt_edges_() %>%
gt_graph() -> g
}
)
test replications elapsed relative user.self sys.self user.child sys.child
1 v3.2 100 6.55 1.492 6.45 0.06 NA NA
2 v4 100 4.39 1.000 4.33 0.05 NA NA