add example compare2

Author: guoyongzhi

Project.toml

@@ -1,7 +1,7 @@
 name = "WordCloud"
 uuid = "6385f0a0-cb03-45b6-9089-4e0acc74b26b"
 authors = ["guoyongzhi <momoshanghan@163.com>"]
-version = "0.6.7"
+version = "0.6.8"
 
 [deps]
 ColorSchemes = "35d6a980-a343-548e-a6ea-1d62b119f2f4"

README.md

@@ -45,7 +45,7 @@ paint(wc, "alice.png", ratio=0.5, background=outline(wc.mask, color="purple", li
 [![gathering](res/gathering.png)](./examples/gathering.jl)  
 *Run the command `runexample(:gathering)` or `showexample(:gathering)` to get the result.* 
 ## Recolor
-[![recolor](res/recolor.png)](./examples/recolor.jl)![recolor](res/butterfly.png)  
+[![recolor](res/recolor.png)](./examples/recolor.jl)  
 *Run the command `runexample(:recolor)` or `showexample(:recolor)` to get the result.* 
 ## Comparison
 [![compare](res/compare.png)](./examples/compare.jl)  

examples/compare2.jl

@@ -0,0 +1,91 @@
+#md# This is a more symmetrical and accurate way to generate comparison wordclouds, but it may be more time consuming.  
+#md# ### Prepare two wordcloud objects
+using WordCloud
+
+stwords = ["us", "will"];
+cs = WordCloud.randomscheme() #:Set1_8#
+as = WordCloud.randomangles() #(0,90,45,-45)#
+dens = 0.5 #not too high
+wca = wordcloud(
+    processtext(open(pkgdir(WordCloud)*"/res/Barack Obama's First Inaugural Address.txt"), stopwords=WordCloud.stopwords_en ∪ stwords), 
+    colors = cs,
+    angles = as,
+    density = dens,
+    run = x->nothing, #turn off the initimage! and placement! in advance
+)
+wcb = wordcloud(
+    processtext(open(pkgdir(WordCloud)*"/res/Donald Trump's Inaugural Address.txt"), stopwords=WordCloud.stopwords_en ∪ stwords),
+    mask = getsvgmask(wca),
+    colors = cs,
+    angles = as,
+    density = dens,
+    run = x->nothing, 
+)
+#md# ### Make the same words the same style
+samewords = getwords(wca) ∩ getwords(wcb)
+println(length(samewords), " same words")
+for w in samewords
+    setcolors!(wcb, w, getcolors(wca, w))
+    setangles!(wcb, w, getangles(wca, w))
+end
+#md# ### Put the same words at same position
+initimages!(wca)
+initimages!(wcb)
+keep(wca, samewords) do
+    placement!(wca)
+    fit!(wca, 1000) #patient=-1 means no teleport; retry=1 means no rescale
+end
+pin(wca, samewords) do
+    placement!(wca) #place other words
+end
+centers = getpositions(wca, samewords, type=getcenter)
+setpositions!(wcb, samewords, centers, type=setcenter!) #manually initialize the position,
+pin(wcb, samewords) do
+    placement!(wcb) #place other words
+end
+#md# ### Fit them all
+function syncposition(samewords, pos, wca, wcb)
+    pos2 = getpositions(wca, samewords, type=getcenter)
+    if pos != pos2
+        setpositions!(wcb, samewords, pos2, type=setcenter!)
+        setstate!(wcb, :placement!)
+    end
+    pos2
+end
+function pinfit!(wc, samewords, ep1, ep2)
+    pin(wc, samewords) do
+        fit!(wc, ep1)
+    end
+    fit!(wc, ep2, patient=-1) #patient=-1 means no teleport
+end
+pos = getpositions(wca, samewords, type=getcenter)
+while wca.params[:epoch] < 2000
+    global pos
+    pinfit!(wca, samewords, 200, 50)
+    pos = syncposition(samewords, pos, wca, wcb)
+    pinfit!(wcb, samewords, 200, 50)
+    pos = syncposition(samewords, pos, wcb, wca)
+    if getstate(wca) == getstate(wcb) == :fit!
+        break
+    end
+end
+println("Takes $(wca.params[:epoch]) and $(wcb.params[:epoch]) epochs")
+WordCloud.printcollisions(wca)
+WordCloud.printcollisions(wcb)
+#md# 
+ma = paint(wca)
+mb = paint(wcb)
+h,w = size(ma)
+space = fill(mb[1], (h, w÷20))
+try mkdir("address_compare2") catch end
+println("results are saved in address_compare2")
+WordCloud.save("address_compare2/compare2.png", [ma space mb])
+#eval# try rm("address_compare2", force=true, recursive=true) catch end 
+gif = WordCloud.GIF("address_compare2")
+record(wca, "Obama", gif)
+record(wcb, "Trump", gif)
+WordCloud.Render.generate(gif, framerate=1)
+wca, wcb
+#eval# runexample(:compare2)
+#md# ![](address_compare2/compare2.png)  
+#md# ![](address_compare2/result.gif)  

src/WordCloud.jl

@@ -3,7 +3,7 @@ export wordcloud, processtext, html2text, countwords, lemmatize, casemerge!
 export parsecolor, rendertext, shape, ellipse, box, loadmask, outline, padding, paint, paintsvg, svgstring
 export imageof, showmask, showmask!
 export record, placement!, rescale!, recolor!, take, keep, ignore, pin, runexample, showexample, 
-    generate!, generate_animation!
+    generate!, generate_animation!, fit!
 export getcolors, getangles, getwords, getweights, setcolors!, setangles!, setwords!, setweights!,
     getpositions, setpositions!, getimages, getsvgimages, setimages!, setsvgimages!, getmask, getsvgmask, 
     getfontsizes, setfontsizes!, getfonts, setfonts!, getstate, setstate!, initimage!, initimages!

src/wc-method.jl

@@ -8,7 +8,7 @@ function initqtree!(wc, i::Integer; backgroundcolor=(0,0,0,0), border=wc.params[
 end
 initqtree!(wc, i; kargs...) = initqtree!.(wc, index(wc, i); kargs...)
 "Initialize word's images and other resources with specified style"
-function initimage!(wc, i::Integer; backgroundcolor=(0,0,0,0), border=wc.params[:border],
+function initimages!(wc, i::Integer; backgroundcolor=(0,0,0,0), border=wc.params[:border],
                     fontsize=getfontsizes(wc, i), color=wc.params[:colors][i],
                     angle = wc.params[:angles][i], font=getfonts(wc, i))
     img, svg = prepareword(wc.words[i], fontsize, color, angle,
@@ -18,8 +18,8 @@ function initimage!(wc, i::Integer; backgroundcolor=(0,0,0,0), border=wc.params[
     initqtree!(wc, i, backgroundcolor=backgroundcolor, border=border)
     nothing
 end
-initimage!(wc, i; kargs...) = initimage!.(wc, index(wc, i); kargs...)
-function initimage!(wc::WC; maxiter=5, error=0.02)
+initimages!(wc, i; kargs...) = initimage!.(wc, index(wc, i); kargs...)
+function initimages!(wc::WC; maxiter=5, error=0.02)
     params = wc.params
     mask = wc.mask
 
@@ -36,10 +36,10 @@ function initimage!(wc::WC; maxiter=5, error=0.02)
     scale = find_weight_scale!(wc, density=params[:density], maxiter=maxiter, error=error)
     println("density set to $(params[:density]), with scale=$scale, font minimum is $(getfontsizes(wc, length(wc.words)))")
     initimage!.(wc, 1:length(words))
-    params[:state] = nameof(initimages!)
+    setstate!(wc, nameof(initimages!))
     wc
 end
-initimages! = initimage!
+initimage! = initimages!
 """
 * placement!(wc)
 * placement!(wc, style=:uniform)
@@ -50,7 +50,7 @@ p=1 produces a rhombus, p=2 produces an ellipse, p>2 produces a rectangle with r
 When you have set `style=:gathering`, you should disable teleporting in `generate!` at the same time(`generate!(wc, patient=-1)`).
 """
 function placement!(wc::WC; style=:uniform, kargs...)
-    if getstate(wc) == nameof(wordcloud)
+    if STATEIDS[getstate(wc)] < STATEIDS[:initimages!]
         initimages!(wc)
     end
     @assert style in [:uniform, :gathering]
@@ -69,7 +69,7 @@ function placement!(wc::WC; style=:uniform, kargs...)
         end
         if ind === nothing error("no room for placement") end
     end
-    wc.params[:state] = nameof(placement!)
+    setstate!(wc, nameof(placement!))
     wc
 end
 
@@ -151,6 +151,30 @@ function recolor!(wc, args...; style=:average, kargs...)
     end
     nothing
 end
+
+function fit!(wc, args...; krags...)
+    if STATEIDS[getstate(wc)] < STATEIDS[:placement!]
+        placement!(wc)
+    end
+    qtrees = [wc.maskqtree, wc.qtrees...]
+    ep, nc = train!(qtrees, args...; krags...)
+    wc.params[:epoch] += ep
+    if nc == 0
+        setstate!(wc, nameof(fit!))
+    end
+    wc
+end
+function printcollisions(wc)
+    qtrees = [wc.maskqtree, wc.qtrees...]
+    colllist = first.(batchcollision(qtrees))
+    get_text(i) = i>1 ? wc.words[i-1] : "#MASK#"
+    collwords = [(get_text(i), get_text(j)) for (i,j) in colllist]
+    if length(colllist) > 0
+        println("have $(length(colllist)) collisions.",
+        " try setting a larger `nepoch` and `retry`, or lower `density` in `wordcloud` to fix that")
+        println("$collwords")
+    end
+end
 """
 # Positional Args
 * wc: the wordcloud to train
@@ -161,10 +185,9 @@ end
 * trainer: appoint a training engine
 """
 function generate!(wc::WC, args...; retry=3, krags...)
-    if getstate(wc) != nameof(placement!) && getstate(wc) != nameof(generate!)
+    if STATEIDS[getstate(wc)] < STATEIDS[:placement!]
         placement!(wc)
     end
-    qtrees = [wc.maskqtree, wc.qtrees...]
     ep, nc = -1, -1
     for r in 1:retry
         if r != 1
@@ -175,28 +198,19 @@ function generate!(wc::WC, args...; retry=3, krags...)
         else
             println("#$r. scale = $(wc.params[:scale])")
         end
-        ep, nc = train!(qtrees, args...; krags...)
-        wc.params[:epoch] += ep
-        if nc == 0
+        fit!(wc, args...; krags...)
+        if getstate(wc) == :fit!
             break
         end
     end
-    println("$ep epochs, $nc collisions")
-    if nc == 0
-        wc.params[:state] = nameof(generate!)
+    if STATEIDS[getstate(wc)] >= STATEIDS[:fit!]
+        println("$(wc.params[:epoch]) epochs")
+        setstate!(wc, nameof(generate!))
         # @assert isempty(outofkernelbounds(wc.maskqtree, wc.qtrees))
         # colllist = first.(batchcollision(qtrees))
         # @assert length(colllist) == 0
     else #check
-        colllist = first.(batchcollision(qtrees))
-        get_text(i) = i>1 ? wc.words[i-1] : "#MASK#"
-        collwords = [(get_text(i), get_text(j)) for (i,j) in colllist]
-        if length(colllist) > 0
-            wc.params[:completed] = false
-            println("have $(length(colllist)) collisions.",
-            " try setting a larger `nepoch` and `retry`, or lower `density` in `wordcloud` to fix that")
-            println("$collwords")
-        end
+        printcollisions(wc)
     end
     wc
 end
@@ -213,6 +227,10 @@ function generate_animation!(wc::WC, args...; outputdir="gifresult", overwrite=o
     re
 end
 
+STATES = nameof.([wordcloud, initimages!, placement!, fit!, generate!])
+STATEIDS = Dict([s=>i for (i,s) in enumerate(STATES)])
+
+
 """
 keep some words and ignore the others, then execute the function. It's the opposite of `ignore`.
 * keep(fun, wc, ws::String) #keep a word