FM analysis support

Dockerfile

@@ -1,14 +1,20 @@
-FROM python:3
+FROM python:3.9.6-bullseye
 
+#Update the image we start from and install all latex stuf and dependencies for mupdf
 RUN apt-get update -y\
  && apt-get install texlive-full -y\
  && apt-get install graphviz -y\
+ && apt-get install libx11-dev mesa-common-dev libgl1-mesa-dev -y\
  && rm -rf /var/lib/apt/lists/*
 
-
+#Copy varylatex files to the image
 COPY . /varylatex
 WORKDIR /varylatex
 
+#Install mupdf 1.18.16 version
+RUN sh mupdf_install.sh
+
+#Install varylatex python dependencies
 RUN pip install -r requirements.txt
 
 VOLUME /varylatex/build

README.md

@@ -31,6 +31,7 @@ You can also run the app within a `Docker` container : to do this, install docke
 sudo docker build -t varylatex .
 # Create the container
 sudo docker run -it --rm -p 5000:5000 varylatex
+# In case you want to rely on a prebuild image there is one at https://hub.docker.com/r/diversolab/varylatex
 ```
 And inside the container you can run
 ```bash

main_optimizer.py

@@ -12,6 +12,8 @@
 from vary.model.generation.generate import generate_random, generate_pdf
 from vary.model.generation.compile import generate_bbl
 from vary.model.decision_trees.analysis import decision_tree
+from vary.model.feature_models.fm_operations import getFeatureHeaders, getProducts, getRandomProduct
+
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()
@@ -31,6 +33,7 @@
                         It needs to have a 'values.json' file and the document must include 'macros' and 'values'")
     parser.add_argument("-c", "--config", help="Generate a specific PDF from a config JSON string")
     parser.add_argument("-p", "--maxpages", type=int, help="The maximum amount of pages accepted for the document")
+    parser.add_argument("-fm","--feature-model", action="store_true", help="This is relying on Famapy to generate the set of configurations. Now this allows contratins. However, a translation to the varylatex format has to be manually done" )
     args = parser.parse_args()
 
     document_path = args.source
@@ -42,18 +45,25 @@
 
     temp_path = create_temporary_copy(document_path)
 
-    conf_source_path = os.path.join(document_path, "variables.json")
-    with open(conf_source_path) as f:
-        conf_source = json.load(f)
+    if args.feature_model :
+        conf_source_path = os.path.join(document_path, "feature_model.uvl")
+        #DataFrame init
+        cols = getFeatureHeaders(conf_source_path)
+        df = pd.DataFrame(columns=cols)
 
-    # DataFrame initialisation 
-    cols = conf_source["booleans"] + \
-        list(conf_source["numbers"].keys()) + \
-        list(conf_source["enums"].keys()) + \
-        list(flatten(conf_source["choices"])) + \
-        ["nbPages", "space"]
-    df = pd.DataFrame(columns=cols)
+    else:
+        conf_source_path = os.path.join(document_path, "variables.json")
+        with open(conf_source_path) as f:
+            conf_source = json.load(f)
 
+        # DataFrame initialisation 
+        cols = conf_source["booleans"] + \
+            list(conf_source["numbers"].keys()) + \
+            list(conf_source["enums"].keys()) + \
+            list(flatten(conf_source["choices"])) + \
+            ["nbPages", "space"]
+        df = pd.DataFrame(columns=cols)
+    print(df)
     file_path = os.path.join(temp_path, filename)
     inject_space_indicator(file_path)
     # LaTeX bbl pregeneration
@@ -67,9 +77,17 @@
         pdf_name = filename+".pdf"
         shutil.copyfile(os.path.join(temp_path, pdf_name), os.path.join(args.output, pdf_name))
     else:
+        if args.feature_model:
+            products = getProducts(conf_source_path)
         for i in range(args.generations):
-            row = generate_random(conf_source, filename, temp_path)
+            if args.feature_model:
+                row = getRandomProduct(products)
+                row = generate_pdf(row, filename, temp_path)
+            else:
+                row = generate_random(conf_source, filename, temp_path)
+
             df = df.append(row, ignore_index=True)
+
             if args.verbose:
                 print(f"Doc {i} generated")
 

mupdf_install.sh

@@ -0,0 +1,7 @@
+#\bin\sh
+wget -O mupdf.tar.gz https://mupdf.com/downloads/archive/mupdf-1.18.0-source.tar.gz ;
+tar -xzf mupdf.tar.gz
+ls
+cd mupdf-1.18.0-source
+make
+make HAVE_X11=no HAVE_GLUT=no prefix=/usr/local install

requirements.txt

@@ -1,6 +1,9 @@
 Flask==1.1.2
 intervaltree==3.0.2
-PyMuPDF==1.17.1
+PyMuPDF==1.18.16
 pandas==1.0.5
 requests==2.22.0
-scikit-learn==0.23.1
+scikit-learn==0.23.1
+famapy==0.5.1
+famapy-fm==0.5.0
+famapy-sat==0.5.1

vary/example/fse/feature_model.uvl

@@ -0,0 +1,38 @@
+imports 
+
+features
+    Paper { abstract }
+        optional
+            PL_FOOTNOTE
+            ACK
+            LONG_AFFILIATION
+            EMAIL
+            LONG_ACK
+        mandatory
+            JS_STYLE { abstract }
+                alternative
+                    tiny
+                    scriptsize
+                    footnotesize
+            CHOICES { abstract }
+                or
+                    PARAGRAPH_ACK
+                    BOLD_ACK
+            vspace_bib { abstract }
+                alternative
+                    vspace_ONE
+                    vspace_FIVE
+                    vspace_TWO
+            bref_size { abstract }
+                alternative
+                    bref_size_POINT_SEVEN
+                    bref_size_ONE
+                    bref_size_TWO
+            cserver_size { abstract }
+                alternative
+                    cserver_size_POINT_SIX
+                    cserver_size_POINT_NINE
+                    cserver_size_TWO
+
+constraints
+

vary/model/feature_models/fm_operations.py

@@ -0,0 +1,93 @@
+from famapy.core.discover import DiscoverMetamodels # This loads the tool in the python execution eviroment
+import random
+
+def transcodeProducts(rawProducts):
+    transcoddedRes=[]
+    for result in rawProducts:
+        r={}
+        if "PL_FOOTNOTE" in result:
+            r["PL_FOOTNOTE"]=True
+        else:
+            r["PL_FOOTNOTE"]=False
+
+        if "ACK" in result:
+            r["ACK"]=True
+        else:
+            r["ACK"]=False
+
+        if "LONG_AFFILIATION" in result:
+            r["LONG_AFFILIATION"]=True
+        else:
+            r["LONG_AFFILIATION"]=False
+
+        if "EMAIL" in result:
+            r["EMAIL"]=True
+        else:
+            r["EMAIL"]=False
+
+        if "LONG_ACK" in result:
+            r["LONG_ACK"]=True
+        else:
+            r["LONG_ACK"]=False
+
+        if "vspace_ONE" in result:
+            r["vspace"]=1
+        elif "vspace_FIVE" in result:
+            r["vspace"]=5
+        elif "vspace_TWO" in result:
+            r["vspace"]=2
+        else:
+            raise Exception
+
+        if "bref_size_POINT_SEVEN" in result:
+            r["bref_size"]=0.7
+        elif "bref_size_ONE" in result:
+            r["bref_size"]=1
+        elif "bref_size_TWO" in result:
+            r["bref_size"]=2
+        else:
+            raise Exception
+
+        if "cserver_size_POINT_SIX" in result:
+            r["cserver_size"]=0.6
+        elif "cserver_size_POINT_NINE" in result:
+            r["cserver_size"]=0.9
+        elif "cserver_size_TWO" in result:
+            r["cserver_size"]=2
+        else:
+            raise Exception
+
+        if "tiny" in result:
+            r["js_style"]="\\tiny"
+        elif "scriptsize" in result:
+            r["js_style"]="\\scriptsize"
+        elif "footnotesize" in result:
+            r["js_style"]="\\footnotesize"
+
+        if "PARAGRAPH_ACK" in result:
+            r["PARAGRAPH_ACK"]=True
+        else:
+            r["PARAGRAPH_ACK"]=False
+
+        if "BOLD_ACK" in result:
+            r["BOLD_ACK"]=True
+        else:
+            r["BOLD_ACK"]=False
+
+        transcoddedRes.append(r)
+    return transcoddedRes
+
+def getProducts(path):
+    dm = DiscoverMetamodels()
+    results = dm.use_operation_from_file("Products", path) # This launch the operation and stores the result on the result variable
+    results = transcodeProducts(results)
+    return results
+
+
+def getRandomProduct(products):
+    n = random.randint(0,len(products)-1)
+    return products[n]
+
+def getFeatureHeaders(path):
+    #This is still hardcoded for the example. TODO See how to generalize
+    return ["PL_FOOTNOTE", "ACK", "LONG_AFFILIATION", "EMAIL", "LONG_ACK", "vspace_bib", "bref_size", "cserver_size", "js_style", "PARAGRAPH_ACK", "BOLD_ACK", "nbPages", "space"]

vary/model/generation/generate.py

@@ -14,7 +14,6 @@
 from vary.model.generation.compile import compile_latex
 from vary.model.generation.analyze_pdf import page_count
 
-
 def random_config(conf_source, fixed_values={}):
     """
     Generates a config dictionnary based on the range of values provided from a dictionnary containing the following keys :
@@ -81,8 +80,7 @@ def generate_random(conf_source, filename, temp_path, fixes_values={}):
     """
     config = random_config(conf_source, fixes_values)
     return generate_pdf(config, filename, temp_path)
-
-
+
 def generate_pdfs(filename, source, output, nb_gens, reset=True, fixed_values = {}):
     """
     Creates as many PDFs as specified with nb_gens, from a random config based on conf_source, and calculate