Skip to content

C
clinicaltrials
Commit 69e7c039 authored by Romain Loth's avatar Romain Loth
Browse files
Options

WIP making facets work in bipartite case + further factorizing 

facets (ie clusters and other searchable/colorable attributes) should work with bipartite case, but it meant factorizing facetsBinning() and updateValueFacets() from commit 71507331
parent d06a320b
Hide whitespace changes
Inline Side-by-side
Showing with 250 additions and 187 deletions
doc/developer_manual.md
View file @ 69e7c039
...@@ -21,11 +21,17 @@ This will still evolve but the main steps for any graph initialization messily u ...@@ -21,11 +21,17 @@ This will still evolve but the main steps for any graph initialization messily u
#### About source data #### About source data
- doc/sem typing: follows the node property "`category`" - doc/sem typing: follows the node property "`type`" or if absent, "`category`"
- if the category name is "document" => catSoc (type 0) - if the category name is "document" => catSoc (type 0)
- if the category name contains the str "term" => catSem (type 1) - if the category name contains the str "term" => catSem (type 1)
- coloration: "`age`" "`growth_rate`" + any attribute of type float or int
- clustering: "`cluster_index`" ou nom figurant dans `TW.nodeClusAtt` - `somenode.attributes`: the `attributes` property is always an object
- any attribute listed in the sourcenode.attributes will be indexed if the TW.scanClusters flag is true
- the mapping from attribute values to matching nodes is in TW.Clusters.aType.anAttr.aValue.map !
- coloration: "`age`" "`growth_rate`" + any attribute of type float or int
- clustering: "`cluster_index`" ou nom figurant dans `TW.nodeClusAtt`
- vocabulary: (en cours) any attribute of type string and where the amount of distinct values is < TW.somesettings
## User interaction mecanisms ## User interaction mecanisms
......
tinawebJS/sigma.parseCustom.js
View file @ 69e7c039
...@@ -262,6 +262,175 @@ function sortNodeTypes(observedTypesDict) { ...@@ -262,6 +262,175 @@ function sortNodeTypes(observedTypesDict) {
return {'categories': observedTypes, 'lookup_dict': catDict} return {'categories': observedTypes, 'lookup_dict': catDict}
} }
// sorting out n.attributes + binning them if too much distinct values
// --------------------------------------------------------------------
// @arg valuesIdx:
// a census of present attribute-values with a 4+ tier structure
// by nodetype
// => by attribute
// => {vals:[allpossiblevalues...],
// map:{eachvalue:[matchingnodeids],
// eachvalue2:[matchingnodeids]...}
// NB vals and map are both useful and complementary
function facetsBinning (valuesIdx, Atts_2_Exclude) {
console.warn("valuesIdx", valuesIdx)
let facetIdx = {}
if (TW.debugFlags.logFacets) {
console.log('dictfyGexf: begin TW.Clusters')
var classvalues_deb = performance.now()
}
// var gotClusters = false
// for (var nodecat in valuesIdx) {
// gotClusters = gotClusters || (valuesIdx[nodecat]['cluster_index'] || valuesIdx[nodecat][TW.nodeClusAtt])
// }
// all scanned attributes get an inverted index
for (var cat in valuesIdx) {
if (!facetIdx[cat]) facetIdx[cat] = {}
for (var at in valuesIdx[cat]) {
// console.log(`======= ${cat}::${at} =======`)
// skip non-numeric or already done
if (Atts_2_Exclude[at] || at == "clust_default") {
continue
}
// array of valueclass/interval/bin objects
facetIdx[cat][at] = []
// if n possible values doesn't need binify
if (Object.keys(valuesIdx[cat][at].map).length <= TW.maxDiscreteValues) {
for (var pval in valuesIdx[cat][at].map) {
facetIdx[cat][at].push({
'labl': `${cat}||${at}||${pval}`,
'val': pval,
// val2ids
'nids': valuesIdx[cat][at].map[pval]
})
}
}
// if binify
else {
var len = valuesIdx[cat][at].vals.length
// sort out vals
valuesIdx[cat][at].vals.sort(function (a,b) {
return Number(a)-Number(b)
})
// (enhanced intervalsInventory)
// => creates bin, binlabels, inverted index per bins
var legendRefTicks = []
// how many bins for this attribute ?
var nBins = 3
if (TW.customLegendsBins && TW.customLegendsBins[at]) {
nBins = TW.customLegendsBins[at]
}
else if (TW.legendsBins) {
nBins = TW.legendsBins
}
// create tick thresholds
for (var l=0 ; l < nBins ; l++) {
let nthVal = Math.floor(len * l / nBins)
legendRefTicks.push(valuesIdx[cat][at].vals[nthVal])
}
if (TW.debugFlags.logFacets) console.debug("intervals for", at, legendRefTicks)
var nTicks = legendRefTicks.length
var sortedDistinctVals = Object.keys(valuesIdx[cat][at].map).sort(function(a,b){return Number(a)-Number(b)})
var nDistinctVals = sortedDistinctVals.length
var lastCursor = 0
// create ticks objects with retrieved full info
for (var l in legendRefTicks) {
l = Number(l)
let lowThres = Number(legendRefTicks[l])
let hiThres = null
if (l < nTicks-1) {
hiThres = Number(legendRefTicks[l+1])
}
else {
hiThres = Infinity
}
var newTick = {
'labl':'',
'nids':[],
'range':[lowThres, hiThres]
}
// 1) union of idmaps
for (var k = lastCursor ; k <= nDistinctVals ; k++) {
var val = Number(sortedDistinctVals[k])
if (val < lowThres) {
console.error("mixup !!", val, lowThres, at)
}
else if ((val >= lowThres) && (val < hiThres)) {
if (!valuesIdx[cat][at].map[val]) {
console.error("unscanned val2ids mapping", val, at)
}
else {
// eg bin2ids map for 2 <= val < 3
// will be U(val2ids maps for 2, 2.1, 2.2,...,2.9)
for (var j in valuesIdx[cat][at].map[val]) {
newTick.nids.push(valuesIdx[cat][at].map[val][j])
}
}
}
// we're over the interval upper bound
// we just need to remember where we were for next interval
else if (val >= hiThres) {
lastCursor = k
break
}
}
// create label
// round %.6f for display
var labLowThres = Math.round(lowThres*1000000)/1000000
var labHiThres = (l==nTicks-1)? '+ ∞' : Math.round(hiThres*1000000)/1000000
newTick.labl = `${cat}||${at}||[${labLowThres} ; ${labHiThres}]`
// save these bins as the cluster index (aka faceting)
if (newTick.nids.length) {
facetIdx[cat][at].push(newTick)
}
}
}
}
// 'clust_default' is an alias to the user-defined default clustering
if (TW.nodeClusAtt != undefined
&& facetIdx[cat][TW.nodeClusAtt] // <= if found in data
&& !facetIdx[cat]['clust_default'] // <= and if an attr named 'clust_default' was not already in data
) {
facetIdx[cat]['clust_default'] = facetIdx[cat][TW.nodeClusAtt]
}
}
if (TW.debugFlags.logFacets) {
var classvalues_fin = performance.now()
console.log('end TW.Clusters, own time:', classvalues_fin-classvalues_deb)
}
return facetIdx
}
// Level-00 // Level-00
// for {1,2}partite graphs // for {1,2}partite graphs
function dictfyGexf( gexf , categories ){ function dictfyGexf( gexf , categories ){
...@@ -382,7 +551,7 @@ function dictfyGexf( gexf , categories ){ ...@@ -382,7 +551,7 @@ function dictfyGexf( gexf , categories ){
var attributes = [] var attributes = []
var attvalueNodes = elNode.getElementsByTagName('attvalue'); var attvalueNodes = elNode.getElementsByTagName('attvalue');
var atts={}; var atts={};
for(k=0; k<attvalueNodes.length; k++){ for(var k=0; k<attvalueNodes.length; k++){
var attvalueNode = attvalueNodes[k]; var attvalueNode = attvalueNodes[k];
var attr = attvalueNode.getAttribute('for'); var attr = attvalueNode.getAttribute('for');
var val = attvalueNode.getAttribute('value'); var val = attvalueNode.getAttribute('value');
...@@ -424,25 +593,9 @@ function dictfyGexf( gexf , categories ){ ...@@ -424,25 +593,9 @@ function dictfyGexf( gexf , categories ){
maxNodeSize= node.size; maxNodeSize= node.size;
// console.debug("node.attributes", node.attributes) // console.debug("node.attributes", node.attributes)
// creating a faceted index from node.attributes
if (TW.scanClusters) { if (TW.scanClusters) {
if (!tmpVals[node_cat]) tmpVals[node_cat]={} [tmpVals, Atts_2_Exclude] = updateValueFacets(tmpVals, Atts_2_Exclude, node)
for (var at in node.attributes) {
if (!tmpVals[node_cat][at]) tmpVals[node_cat][at]={'vals':[],'map':{}}
let someval = Number(node.attributes[at])
// Identifying the attribute datatype: exclude strings and objects
if ( isNaN(someval) ) {
if (!Atts_2_Exclude[at]) Atts_2_Exclude[at]=true;
}
// numeric attr => build facets
else {
if (!tmpVals[node_cat][at].map[someval]) tmpVals[node_cat][at].map[someval] = []
tmpVals[node_cat][at].vals.push(someval) // for ordered scale
tmpVals[node_cat][at].map[someval].push(node.id) // inverted index
}
}
} }
} // finish nodes loop } // finish nodes loop
...@@ -461,160 +614,10 @@ function dictfyGexf( gexf , categories ){ ...@@ -461,160 +614,10 @@ function dictfyGexf( gexf , categories ){
// ------------- /debug: for local stats ---------------- // ------------- /debug: for local stats ----------------
if (TW.debugFlags.logFacets) {
console.log('dictfyGexf: begin TW.Clusters')
var classvalues_deb = performance.now()
}
var gotClusters = false
for (var nodecat in tmpVals) {
gotClusters = gotClusters || (tmpVals[nodecat]['cluster_index'] || tmpVals[nodecat][TW.nodeClusAtt])
}
// clusters and other facets => type => name => [{label,val/range,nodeids}] // clusters and other facets => type => name => [{label,val/range,nodeids}]
TW.Clusters = {} TW.Clusters = facetsBinning(tmpVals, Atts_2_Exclude)
// sorting out properties in n.attributes ==> £TODO shared function createClusterIndex() up to classvalues_fin
// --------------------------------------
// all scanned attributes get an inverted index
for (var cat in tmpVals) {
if (!TW.Clusters[cat]) TW.Clusters[cat] = {}
for (var at in tmpVals[cat]) {
// console.log(`======= ${cat}::${at} =======`)
var allTicks = []
// skip non-numeric or already done
if (Atts_2_Exclude[at] || at == "clust_default") {
continue
}
// array of valueclass/interval/bin objects
TW.Clusters[cat][at] = []
// if n possible values doesn't need binify
if (Object.keys(tmpVals[cat][at].map).length <= TW.maxDiscreteValues) {
for (var pval in tmpVals[cat][at].map) {
TW.Clusters[cat][at].push({
'labl': `${cat}||${at}||${pval}`,
'val': pval,
// val2ids
'nids': tmpVals[cat][at].map[pval]
})
}
}
// if binify
else {
var len = tmpVals[cat][at].vals.length
// sort out vals
tmpVals[cat][at].vals.sort(function (a,b) {
return Number(a)-Number(b)
})
// (enhanced intervalsInventory)
// => creates bin, binlabels, inverted index per bins
var legendRefTicks = []
// how many bins for this attribute ?
var nBins = 3
if (TW.customLegendsBins && TW.customLegendsBins[at]) {
nBins = TW.customLegendsBins[at]
}
else if (TW.legendsBins) {
nBins = TW.legendsBins
}
// create tick thresholds
for (var l=0 ; l < nBins ; l++) {
let nthVal = Math.floor(len * l / nBins)
legendRefTicks.push(tmpVals[cat][at].vals[nthVal])
}
if (TW.debugFlags.logFacets) console.debug("intervals for", at, legendRefTicks)
var nTicks = legendRefTicks.length
var sortedDistinctVals = Object.keys(tmpVals[cat][at].map).sort(function(a,b){return Number(a)-Number(b)})
var nDistinctVals = sortedDistinctVals.length
var lastCursor = 0
// create ticks objects with retrieved full info
for (var l in legendRefTicks) {
l = Number(l)
let lowThres = Number(legendRefTicks[l])
let hiThres = null
if (l < nTicks-1) {
hiThres = Number(legendRefTicks[l+1])
}
else {
hiThres = Infinity
}
var newTick = {
'labl':'',
'nids':[],
'range':[lowThres, hiThres]
}
// 1) union of idmaps
for (var k = lastCursor ; k <= nDistinctVals ; k++) {
var val = Number(sortedDistinctVals[k])
if (val < lowThres) {
console.error("mixup !!", val, lowThres, at)
}
else if ((val >= lowThres) && (val < hiThres)) {
if (!tmpVals[cat][at].map[val]) {
console.error("unscanned val2ids mapping", val, at)
}
else {
// eg bin2ids map for 2 <= val < 3
// will be U(val2ids maps for 2, 2.1, 2.2,...,2.9)
for (var j in tmpVals[cat][at].map[val]) {
newTick.nids.push(tmpVals[cat][at].map[val][j])
}
}
}
// we're over the interval upper bound
// we just need to remember where we were for next interval
else if (val >= hiThres) {
lastCursor = k
break
}
}
// create label
// round %.6f for display
var labLowThres = Math.round(lowThres*1000000)/1000000
var labHiThres = (l==nTicks-1)? '+ ∞' : Math.round(hiThres*1000000)/1000000
newTick.labl = `${cat}||${at}||[${labLowThres} ; ${labHiThres}]`
// save these bins as the cluster index (aka faceting)
if (newTick.nids.length) {
TW.Clusters[cat][at].push(newTick)
}
}
}
}
// 'clust_default' is an alias to the user-defined default clustering
if (TW.nodeClusAtt != undefined
&& TW.Clusters[cat][TW.nodeClusAtt] // <= if found in data
&& !TW.Clusters[cat]['clust_default'] // <= and if an attr named 'clust_default' was not already in data
) {
TW.Clusters[cat]['clust_default'] = TW.Clusters[cat][TW.nodeClusAtt]
}
}
if (TW.debugFlags.logFacets) {
var classvalues_fin = performance.now()
console.log('end TW.Clusters, own time:', classvalues_fin-classvalues_deb)
}
//New scale for node size: now, between 2 and 5 instead [1,70] //New scale for node size: now, between 2 and 5 instead [1,70]
...@@ -745,6 +748,37 @@ function updateRelations(typedRelations, edgeCateg, srcId, tgtId){ ...@@ -745,6 +748,37 @@ function updateRelations(typedRelations, edgeCateg, srcId, tgtId){
} }
// To fill the reverse map: values => nodeids of a given type
function updateValueFacets(facetIdx, Atts_2_Exclude, aNode){
if (!facetIdx[aNode.type]) facetIdx[aNode.type]={}
for (var at in aNode.attributes) {
if (!facetIdx[aNode.type][at]) facetIdx[aNode.type][at]={'vals':[],'map':{}}
let castVal = Number(aNode.attributes[at])
// Identifying the attribute datatype: exclude strings and objects
if ( isNaN(castVal) ) {
if (!Atts_2_Exclude[at]) Atts_2_Exclude[at]=true;
// TODO: this old Atts_2_Exclude strategy could be replaced,
// not to exclude but to store the datatype somewhere like facetIdx[aNode.type][at].dtype
// => the datatype would be a condition (no bins if not numeric, etc.)
// => it would also allow to index text values (eg country, affiliation, etc.)
// with the strategy "most frequent distinct values" + "others"
// which would be useful (eg country, affiliation, etc.) !!!
}
// numeric attr => build facets
else {
if (!facetIdx[aNode.type][at].map[castVal]) facetIdx[aNode.type][at].map[castVal] = []
facetIdx[aNode.type][at].vals.push(castVal) // for ordered scale
facetIdx[aNode.type][at].map[castVal].push(aNode.id) // inverted index
}
}
return [facetIdx, Atts_2_Exclude]
}
// creates and updates nodes1 nodes2 and bipartiteN2D and bipartiteD2N // creates and updates nodes1 nodes2 and bipartiteN2D and bipartiteD2N
// but seems useless because all info is already in each nodes.type and edge.categ // but seems useless because all info is already in each nodes.type and edge.categ
// (especially when changeType uses a loop on all nodes anyway) // (especially when changeType uses a loop on all nodes anyway)
...@@ -873,6 +907,10 @@ function dictfyJSON( data , categories ) { ...@@ -873,6 +907,10 @@ function dictfyJSON( data , categories ) {
nodes2={}, bipartiteD2N={}, bipartiteN2D={} nodes2={}, bipartiteD2N={}, bipartiteN2D={}
} }
// if scanClusters, we'll also use:
var tmpVals = {}
var Atts_2_Exclude = {}
for(var nid in data.nodes) { for(var nid in data.nodes) {
let n = data.nodes[nid]; let n = data.nodes[nid];
let node = {} let node = {}
...@@ -901,8 +939,17 @@ function dictfyJSON( data , categories ) { ...@@ -901,8 +939,17 @@ function dictfyJSON( data , categories ) {
catCount[node.type]++; catCount[node.type]++;
nodes[node.id] = node; nodes[node.id] = node;
console.log(n.attributes)
// creating a faceted index from node.attributes
if (TW.scanClusters) {
[tmpVals, Atts_2_Exclude] = updateValueFacets(tmpVals, Atts_2_Exclude, node)
}
} }
TW.Clusters = facetsBinning (tmpVals, Atts_2_Exclude)
colorList.sort(function(){ return Math.random()-0.5; }); colorList.sort(function(){ return Math.random()-0.5; });
for (var i in nodes ){ for (var i in nodes ){
if (nodes[i].color=="#FFFFFF") { if (nodes[i].color=="#FFFFFF") {
......
tinawebJS/sigmaUtils.js
View file @ 69e7c039
...@@ -816,17 +816,24 @@ function clustersBy(daclass) { ...@@ -816,17 +816,24 @@ function clustersBy(daclass) {
for(var nid in NodeID_Val) { for(var nid in NodeID_Val) {
var newval_color = Math.round( ( Min_color+(NodeID_Val[nid]["round"]-real_min)*((Max_color-Min_color)/(real_max-real_min)) ) ); var newval_color = Math.round( ( Min_color+(NodeID_Val[nid]["round"]-real_min)*((Max_color-Min_color)/(real_max-real_min)) ) );
var hex_color = rgbToHex(255, (255-newval_color) , 0) var hex_color = rgbToHex(255, (255-newval_color) , 0)
TW.partialGraph.graph.nodes(nid).color = hex_color
TW.partialGraph.graph.nodes(nid).customAttrs.alt_color = hex_color
// FIXME not used ? let n = TW.partialGraph.graph.nodes(nid)
TW.partialGraph.graph.nodes(nid).customAttrs.altgrey_color = false if (n && !n.hidden) {
n.color = hex_color
n.customAttrs.alt_color = hex_color
// FIXME not used ?
n.customAttrs.altgrey_color = false
// £TODO SETTING SIZE HERE SHOULD BE OPTIONAL
var newval_size = Math.round( ( Min_size+(NodeID_Val[nid]["round"]-real_min)*((Max_size-Min_size)/(real_max-real_min)) ) );
n.size = newval_size;
n.label = "("+NodeID_Val[nid]["real"].toFixed(min_pow)+") "+TW.Nodes[nid].label
}
var newval_size = Math.round( ( Min_size+(NodeID_Val[nid]["round"]-real_min)*((Max_size-Min_size)/(real_max-real_min)) ) );
TW.partialGraph.graph.nodes(nid).size = newval_size;
// console.log("real:"+ NodeID_Val[i]["real"] + " | newvalue: "+newval_size) // console.log("real:"+ NodeID_Val[i]["real"] + " | newvalue: "+newval_size)
TW.partialGraph.graph.nodes(nid).label = "("+NodeID_Val[nid]["real"].toFixed(min_pow)+") "+TW.Nodes[nid].label
} }
// [ / Scaling node colours(0-255) and sizes(2-7) ] // [ / Scaling node colours(0-255) and sizes(2-7) ]
...@@ -855,15 +862,18 @@ function repaintEdges() { ...@@ -855,15 +862,18 @@ function repaintEdges() {
} }
else { else {
let e = TW.partialGraph.graph.edges(eid) let e = TW.partialGraph.graph.edges(eid)
let src = TW.partialGraph.graph.nodes(idPair[0])
let tgt = TW.partialGraph.graph.nodes(idPair[1]) if (e) {
let src = TW.partialGraph.graph.nodes(idPair[0])
let tgt = TW.partialGraph.graph.nodes(idPair[1])
if (src && tgt) { if (src && tgt) {
let src_color = src.customAttrs.alt_color || '#555' let src_color = src.customAttrs.alt_color || '#555'
let tgt_color = tgt.customAttrs.alt_color || '#555' let tgt_color = tgt.customAttrs.alt_color || '#555'
e.customAttrs.alt_rgb = sigmaTools.edgeRGB(src_color,tgt_color) e.customAttrs.alt_rgb = sigmaTools.edgeRGB(src_color,tgt_color)
// we don't set e.color because opacity may vary if selected or not // we don't set e.color because opacity may vary if selected or not
}
} }
} }
} }
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment