Out of curiosity, is there any public stat on discourse activity? Like posts or new users / month or stuff like that?
Hereās the data in graphic form
and hereās some code if you want to play around yourself
using CSV, DataFrames, DSP, Plots
default(show=false)
data = CSV.read("/home/fredrikb/Downloads/daily_activity.csv", DataFrame)
sort!(data, :date)
data = data[200:end-1, :] # remove the very beginning and the current day
##
function mf(x)
m = mean(x)
# filtfilt(df, x .- m) .+ m
filtfilt(ones(14), [14], x .- m) .+ m
end
fields = ["users", "posts", "pageviews", "likes"]
figs = map(fields) do field
x = data[!, field]
x_filt = mf(x)
fu1 = plot(data.date, [x x_filt], layout=2, sp=1, title="Daily new $field", linewidth=[1 5], lab=["Daily" "2 week mean"])
plot!(data.date, cumsum(x), sp=2, title="Cumulative $field", lab="")
end
plot(figs..., layout=(length(fields),1)) |> display
Could the trust level to access the stats be lowered to āmemberā (if not making the stats completely visible to everyone) ? There could be some sensitive data on these stats ?
In any event hereās an update:
Note Iām not a signal processing expert like Fredrik so this is a more pedestrian 2-week running mean:
julia> using CSV, DataFrames, Plots, RollingFunctions
julia> data = sort(CSV.read("discourse.csv", DataFrame), :date)[200:end-1, :];
julia> ps = []
Any[]
julia> for (v, t) ā zip(["users", "posts", "pageviews", "likes"], ["Daily new users", "Daily new posts", "Daily new pageviews", "Daily new likes"])
p = plot(data.date, data[!, v], label = "Daily", lw = 1, alpha = 0.5, title = t)
plot!(p, data.date, runmean(data[!, v], 14), label = "14-day average")
push!(ps, p)
end
julia> plot(ps..., size = (800, 700), xrot = 75)
Decreasing trend in likes per day suggests @Tamas_Papp has been less active over the last yearā¦
The one metric missing from the original post and my post above is topics, so here for completeness:
Indeed it seems like likes have declined the most out of all metrics, which is probably the Tamas effect. At the same time pageviews is the only metric with an entirely unbroken upward trend, so maybe thereās more and more people out there who find Discourse through a web search and get the info they need without ever having to sign up, post, or like anything.
Thereās no sensitive information here ā if there were we wouldnāt make it available to anyone. It does, however, make sense to limit its access for both community and server load reasons. Itās an expensive operation that we donāt want lots of folks to run, and generally the folks who are most curious are those who are most invested in the community.
Edit: there are some basic statistics available in the about page: About - Julia Programming Language