func run() int {
args, optargs, err := getopt.GetOpt(
os.Args[1:],
"h:p:n:",
[]string{
"help",
"pattern=",
"cpu-profile=",
"names=",
},
)
if err != nil {
fmt.Fprintln(os.Stderr, err)
cmd.Usage(cmd.ErrorCodes["opts"])
}
patterns := make([]string, 0, 10)
namesPath := ""
cpuProfile := ""
for _, oa := range optargs {
switch oa.Opt() {
case "-h", "--help":
cmd.Usage(0)
case "-p", "--pattern":
patterns = append(patterns, oa.Arg())
case "-n", "--names":
namesPath = cmd.AssertFileExists(oa.Arg())
case "--cpu-profile":
cpuProfile = cmd.AssertFile(oa.Arg())
default:
fmt.Fprintf(os.Stderr, "Unknown flag '%v'\n", oa.Opt())
cmd.Usage(cmd.ErrorCodes["opts"])
}
}
if namesPath != "" && len(patterns) > 0 {
fmt.Fprintf(os.Stderr, "You cannot supply patterns with both (-p) and (-n)\n")
cmd.Usage(cmd.ErrorCodes["opts"])
}
if len(patterns) == 0 && namesPath == "" {
fmt.Fprintf(os.Stderr, "You must supply a pattern (-p, -n)\n")
cmd.Usage(cmd.ErrorCodes["opts"])
}
if namesPath != "" {
var err error
patterns, _, err = loadNames(namesPath)
if err != nil {
fmt.Fprintf(os.Stderr, "There was error loading the probability file\n")
fmt.Fprintf(os.Stderr, "%v\n", err)
return 1
}
}
conf := &config.Config{}
graphs := make([]*digraph.Digraph, 0, 10)
for len(args) > 0 {
loadDt, as := cmd.ParseType(args, conf)
args = as
dt, _ := loadDt(nil)
graph := dt.(*digraph.Digraph)
graphs = append(graphs, graph)
}
if cpuProfile != "" {
defer cmd.CPUProfile(cpuProfile)()
}
errors.Logf("INFO", "looking for embeddings")
for _, graph := range graphs {
for _, pattern := range patterns {
sg, err := subgraph.ParsePretty(pattern, graph.Labels)
if err != nil {
fmt.Fprintf(os.Stderr, "There was error during the parsing the pattern '%v'\n", pattern)
fmt.Fprintf(os.Stderr, "%v\n", err)
return 1
}
if sg.Pretty(graph.Labels) != pattern {
errors.Logf("ERROR", "bad load of pattern")
errors.Logf("ERROR", "expected %v", pattern)
errors.Logf("ERROR", "got %v", sg.Pretty(graph.Labels))
return 1
}
errors.Logf("INFO", "cur sg: %v", sg.Pretty(graph.Labels))
ei, _ := sg.IterEmbeddings(subgraph.MostConnected, graph.Indices, nil, nil, nil)
c := 0
for _, next := ei(false); next != nil; _, next = next(false) {
c++
}
errors.Logf("EMB", "total embs: %v", c)
fmt.Println(sg.Dotty(graph.Labels, nil, nil))
}
}
return 0
}
func run() int {
args, optargs, err := getopt.GetOpt(
os.Args[1:],
"h:p:v:",
[]string{
"help",
"pattern=",
"cpu-profile=",
"visualize=",
"probabilities=",
"samples=",
"names=",
},
)
if err != nil {
fmt.Fprintln(os.Stderr, err)
cmd.Usage(cmd.ErrorCodes["opts"])
}
visual := ""
patterns := make([]string, 0, 10)
prPath := ""
namesPath := ""
cpuProfile := ""
samples := -1
for _, oa := range optargs {
switch oa.Opt() {
case "-h", "--help":
cmd.Usage(0)
case "-p", "--pattern":
patterns = append(patterns, oa.Arg())
case "--probabilities":
prPath = cmd.AssertFileExists(oa.Arg())
case "--names":
namesPath = cmd.AssertFileExists(oa.Arg())
case "--samples":
samples = cmd.ParseInt(oa.Arg())
case "--cpu-profile":
cpuProfile = cmd.AssertFile(oa.Arg())
case "-v", "--visual":
visual = cmd.AssertFile(oa.Arg())
default:
fmt.Fprintf(os.Stderr, "Unknown flag '%v'\n", oa.Opt())
cmd.Usage(cmd.ErrorCodes["opts"])
}
}
if (prPath != "" || namesPath != "") && len(patterns) > 0 {
fmt.Fprintf(os.Stderr, "You cannot supply patterns with both (-p) and (--probabilities)\n")
cmd.Usage(cmd.ErrorCodes["opts"])
}
if len(patterns) == 0 && prPath == "" && namesPath == "" {
fmt.Fprintf(os.Stderr, "You must supply a pattern (-p, --names, --probabilities)\n")
cmd.Usage(cmd.ErrorCodes["opts"])
}
var patternCount int = len(patterns)
var prs []float64 = nil
if prPath != "" {
var err error
prs, patterns, patternCount, err = loadProbabilities(prPath)
if err != nil {
fmt.Fprintf(os.Stderr, "There was error loading the probability file\n")
fmt.Fprintf(os.Stderr, "%v\n", err)
return 1
}
} else if namesPath != "" {
var err error
prs, patterns, patternCount, err = loadNames(namesPath)
if err != nil {
fmt.Fprintf(os.Stderr, "There was error loading the probability file\n")
fmt.Fprintf(os.Stderr, "%v\n", err)
return 1
}
}
if samples < len(prs) {
errors.Errorf("INFO", "assuming # of samples is the total number of patterns supplied: %v", patternCount)
samples = patternCount
}
conf := &config.Config{}
graphs := make([]*digraph.Digraph, 0, 10)
for len(args) > 0 {
loadDt, as := cmd.ParseType(args, conf)
args = as
dt, _ := loadDt(nil)
graph := dt.(*digraph.Digraph)
graphs = append(graphs, graph)
}
if cpuProfile != "" {
defer cmd.CPUProfile(cpuProfile)()
}
var visualize io.Writer = nil
if visual != "" {
f, err := os.Create(visual)
if err != nil {
fmt.Fprintf(os.Stderr, "There was error opening the visualization output file\n")
fmt.Fprintf(os.Stderr, "%v\n", err)
return 1
}
defer f.Close()
visualize = f
}
matches := make([]float64, 0, len(patterns))
matched := make([]*subgraph.SubGraph, 0, len(patterns))
sgEdges := make([]float64, 0, len(patterns))
total := 0.0
totalEdges := 0.0
for _, graph := range graphs {
for _, pattern := range patterns {
sg, err := subgraph.ParsePretty(pattern, graph.Labels)
if err != nil {
fmt.Fprintf(os.Stderr, "There was error during the parsing the pattern '%v'\n", pattern)
fmt.Fprintf(os.Stderr, "%v\n", err)
return 1
}
match, csg, err := sg.EstimateMatch(graph.Indices)
match = match * float64(len(sg.E))
if err != nil {
errors.Logf("ERROR", "%v", err)
return 1
}
matches = append(matches, match)
matched = append(matched, csg)
sgEdges = append(sgEdges, float64(len(sg.E)))
// fmt.Printf("%v, %v, %v\n", i+1, match, pattern)
total += match
totalEdges += float64(len(sg.E))
if visualize != nil {
dotty, err := csg.VisualizeEmbedding(sg.AsIndices(), graph.Labels)
if err != nil {
fmt.Fprintf(os.Stderr, "There was error visualizing the embedding '%v'\n", csg)
fmt.Fprintf(os.Stderr, "%v\n", err)
return 1
}
fmt.Fprintln(visualize, dotty)
}
}
}
errors.Logf("DEBUG", "prs %v", sum(prs))
fmt.Printf(", %v, sample total covered edges\n", total)
fmt.Printf(", %v, sample total edges\n", totalEdges)
fmt.Printf(", %v, sample covered/total\n", total/totalEdges)
fmt.Printf(", %v, sample avg covered\n", total/float64(len(patterns)))
fmt.Printf(", %v, sample avg edges\n", totalEdges/float64(len(patterns)))
if len(prs) > 0 {
pis := samplingPrs(samples, prs)
jpis := jointSamplingPrs(samples, prs, pis)
estN := estPopSize(pis)
estTotalMatch := estPopTotal(pis, matches)
estVarTotalMatch := estVarTotal(pis, jpis, matches)
estTotalEdges := estPopTotal(pis, sgEdges)
estVarTotalEdges := estVarTotal(pis, jpis, sgEdges)
fmt.Printf("\n")
fmt.Printf(", %v, estimated population total of matched edges\n", estTotalMatch)
fmt.Printf(", %v, estimated population total of total edges\n", estTotalEdges)
fmt.Printf(", %v, estimated var population total of match edges\n", estVarTotalMatch)
fmt.Printf(", %v, estimated var population total of total edges\n", estVarTotalEdges)
fmt.Printf(", %v, estimated std population total of match edges\n", math.Sqrt(estVarTotalMatch))
fmt.Printf(", %v, estimated std population total of total edges\n", math.Sqrt(estVarTotalEdges))
fmt.Printf(", %v, estimated population mean\n", estTotalMatch/estTotalEdges)
estMeanMatch := estPopMean(estTotalMatch, estN)
estMeanEdges := estPopMean(estTotalEdges, estN)
fmt.Printf("\n")
fmt.Printf(", %v, est. mean matches\n", estMeanMatch)
fmt.Printf(", %v, est. mean edges\n", estMeanEdges)
fmt.Printf(", %v, est. cover\n", estMeanMatch/estMeanEdges)
varMeanMatch := estVarMean(estN, estMeanMatch, pis, jpis, matches)
varMeanEdges := estVarMean(estN, estMeanEdges, pis, jpis, sgEdges)
stdMeanMatch := math.Sqrt(varMeanMatch)
stdMeanEdges := math.Sqrt(varMeanEdges)
fmt.Printf("\n")
fmt.Printf(", %v, var. mean matches\n", varMeanMatch)
fmt.Printf(", %v, var. mean edges\n", varMeanEdges)
fmt.Printf(", %v, std. mean matches\n", stdMeanMatch)
fmt.Printf(", %v, std. mean edges\n", stdMeanEdges)
t := t_alpha_05[samples-1]
fmt.Printf("\n")
fmt.Printf(", %v - %v, interval. mean matches\n",
estMeanMatch-t*stdMeanMatch,
estMeanMatch+t*stdMeanMatch)
fmt.Printf(", %v - %v, interval. mean edges\n",
estMeanEdges-t*stdMeanEdges,
estMeanEdges+t*stdMeanEdges)
fmt.Printf(", %v - %v, interval. cover\n",
math.Max((estMeanMatch-t*stdMeanMatch)/(estMeanEdges+t*stdMeanEdges), 0.0),
math.Min((estMeanMatch+t*stdMeanMatch)/(estMeanEdges-t*stdMeanEdges), 1.0))
}
return 0
}