DailyMed NDC to Label Image Mart

airflow/dags/dailymed/dag.py

@@ -0,0 +1,149 @@
+from pathlib import Path
+import pendulum
+import zipfile
+import os
+
+from airflow.decorators import dag, task
+from airflow.hooks.subprocess import SubprocessHook
+
+from lxml import etree
+
+from sagerx import create_path, load_df_to_pg
+
+@dag(
+    schedule="0 0 10 * *",
+    start_date=pendulum.yesterday(),
+    catchup=False
+)
+def dailymed():
+    dag_id = "dailymed"
+
+    ds_folder = Path("/opt/airflow/dags") / dag_id
+    data_folder = Path("/opt/airflow/data") / dag_id
+
+    # NOTE: "dm_spl_release_human" accounts for both 
+    # rx and otc SPLs (but no other types of SPLs)
+    # - "dm_spl_release_human_rx" for rx meds only
+    # - "dm_spl_release_human_otc" for otc meds only
+    # - "dm_spl_release_human_rx_part1" for a given part
+    # - "dm_spl_daily_update_MMDDYYYY" for a given date
+    #   (replace MMDDYYY with your month, day, and year)
+    file_set = "dm_spl_release_human_rx"
+
+    def connect_to_ftp_dir(ftp_str: str, dir: str):
+        import ftplib
+
+        ftp = ftplib.FTP(ftp_str)
+        ftp.login()
+
+        ftp.cwd(dir)
+
+        return ftp
+
+    def obtain_ftp_file_list(ftp):
+        import fnmatch
+
+        file_list = []
+        for file in ftp.nlst():
+            if fnmatch.fnmatch(file, f"*{file_set}*"):
+                file_list.append(file)
+        return file_list
+
+    def get_dailymed_files(ftp, file_name: str):
+        zip_path = create_path(data_folder) / file_name
+
+        with open(zip_path, "wb") as file:
+            ftp.retrbinary(f"RETR {file_name}", file.write)
+
+        with zipfile.ZipFile(zip_path, "r") as zip_ref:
+            zip_ref.extractall(data_folder.with_suffix(""))
+
+        os.remove(zip_path)
+
+    def transform_xml(input_xml, xslt):
+        # load xml input
+        dom = etree.parse(input_xml, etree.XMLParser(huge_tree=True))
+        # load XSLT
+        xslt_doc = etree.parse(xslt)
+        xslt_transformer = etree.XSLT(xslt_doc)
+        # apply XSLT on loaded dom
+        new_xml = xslt_transformer(dom)
+        return etree.tostring(new_xml, pretty_print=True).decode("utf-8")
+
+    def load_xml_data(spl_type_data_folder: Path):
+        import re
+        import pandas as pd
+        import sqlalchemy
+
+        xslt = ds_folder / "template.xsl"
+
+        db_conn_string = os.environ["AIRFLOW_CONN_POSTGRES_DEFAULT"]
+        db_conn = sqlalchemy.create_engine(db_conn_string)
+
+        data = []
+        for zip_folder in spl_type_data_folder.iterdir():
+            with zipfile.ZipFile(zip_folder) as unzipped_folder:
+                zip_file = zip_folder.stem
+                set_id = zip_file.split('_')[1]
+                for subfile in unzipped_folder.infolist():
+                    if re.search("\.xml$", subfile.filename):
+                        xml_file = subfile.filename
+
+                        # xslt transform
+                        temp_xml_file = unzipped_folder.extract(subfile, spl_type_data_folder)
+                        xml_content = transform_xml(temp_xml_file, xslt)
+                        os.remove(temp_xml_file)
+
+                        # append row to the data list
+                        data.append({"set_id": set_id, "zip_file": zip_file, "xml_file": xml_file, "xml_content": xml_content})
+
+        df = pd.DataFrame(
+            data,
+            columns=["set_id", "zip_file", "xml_file", "xml_content"],
+        )
+
+        load_df_to_pg(
+            df,
+            schema_name="sagerx_lake",
+            table_name="dailymed",
+            if_exists="append", # TODO: make this better - maybe don't put stuff in multiple folders?
+            index=False,
+        )
+
+    @task
+    def extract():
+        dailymed_ftp = "public.nlm.nih.gov"
+        ftp_dir = "/nlmdata/.dailymed/"
+
+        ftp = connect_to_ftp_dir(dailymed_ftp, ftp_dir)
+
+        file_list = obtain_ftp_file_list(ftp)
+        print(f'Extracting {file_list}')
+
+        for file_name in file_list:
+            get_dailymed_files(ftp, file_name)
+
+    @task
+    def load():
+        spl_types = ['prescription', 'otc']
+
+        for spl_type in spl_types:
+            spl_type_data_folder = (
+                data_folder
+                / spl_type
+            )
+            if os.path.exists(spl_type_data_folder):
+                print(f'Loading {spl_type} SPLs...')
+                load_xml_data(spl_type_data_folder)
+
+
+    # Task to transform data using dbt
+    @task
+    def transform():
+        subprocess = SubprocessHook()
+        result = subprocess.run_command(['dbt', 'run', '--select', 'models/staging/dailymed', 'models/intermediate/dailymed'], cwd='/dbt/sagerx')
+        print("Result from dbt:", result)
+
+    extract() >> load() >> transform()
+
+dailymed()

airflow/dags/dailymed_daily/dailymed_prescription.xsl → airflow/dags/dailymed/template.xsl

@@ -13,6 +13,10 @@
     <dailymed>
         <xsl:apply-templates select="v3:document"/>
 
+        <PackageLabels>
+            <xsl:apply-templates select="//v3:section[v3:code[@code='51945-4']]"/>
+        </PackageLabels>
+
         <NDCList>
             <xsl:apply-templates select="v3:document/v3:component"/>
         </NDCList>
@@ -67,6 +71,49 @@
     <ApplicationNumber><xsl:value-of select="//v3:subjectOf/v3:approval/v3:id/@extension"/></ApplicationNumber>
 </xsl:template>
 
+<!-- PackageLabels -->
+<xsl:template match="//v3:section[v3:code[@code='51945-4']]">
+    <PackageLabel> <!-- there can be multiple PRINCIPAL DISPLAY PANEL sections in a SPL -->
+        <MediaList>
+            <!-- If v3:observationMedia exists, process only those -->
+            <xsl:choose>
+                <xsl:when test=".//v3:observationMedia">
+                    <xsl:for-each select=".//v3:observationMedia">
+                        <Media>
+                            <ID>
+                                <xsl:value-of select="./@ID"/>
+                            </ID>
+                            <Image>
+                                <xsl:value-of select="v3:value/v3:reference/@value"/>
+                            </Image>
+                        </Media>
+                    </xsl:for-each>
+                </xsl:when>
+                <!-- Else, process v3:renderMultiMedia -->
+                <xsl:otherwise>
+                    <xsl:for-each select=".//v3:renderMultiMedia">
+                        <xsl:variable name="refID" select="@referencedObject"/>
+
+                        <!-- look for the corresponding observationMedia with the same ID -->
+                        <xsl:for-each select="//v3:observationMedia[@ID=$refID]">
+                            <Media>
+                                <ID>
+                                    <xsl:value-of select="./@ID"/>
+                                </ID>
+                                <Image>
+                                    <xsl:value-of select="v3:value/v3:reference/@value"/>
+                                </Image>
+                            </Media>
+                        </xsl:for-each>
+                    </xsl:for-each>
+                </xsl:otherwise>
+            </xsl:choose>
+        </MediaList>
+        <ID><xsl:value-of select=".//@root"/></ID>
+        <Text><xsl:value-of select="."/></Text>
+    </PackageLabel>
+</xsl:template>
+
 <!-- NDCList -->
 <xsl:template match="v3:document/v3:component">
     <xsl:for-each select="//v3:containerPackagedProduct/v3:code[@codeSystem='2.16.840.1.113883.6.69']">