Web indexing: add a parameter to specify at what frequency a page should be stored again: none/day/week/month/year. Previously, only one instance could be kept, which is still the default

src/index/webqueue.cpp

@@ -173,14 +173,25 @@ public:
 
 // Initialize. Compute paths and create a temporary directory that will be
 // used by internfile()
-WebQueueIndexer::WebQueueIndexer(RclConfig *cnf, Rcl::Db *db,
-                                 DbIxStatusUpdater *updfunc)
-    : m_config(cnf), m_db(db), m_cache(0), m_updater(updfunc), 
-      m_nocacheindex(false)
+WebQueueIndexer::WebQueueIndexer(RclConfig *cnf, Rcl::Db *db, DbIxStatusUpdater *updfunc)
+    : m_config(cnf), m_db(db), m_updater(updfunc)
 {
     m_queuedir = m_config->getWebQueueDir();
     path_catslash(m_queuedir);
     m_cache = new WebStore(cnf);
+    string keepinterval;
+    m_config->getConfParam("webcachekeepinterval", keepinterval);
+    if (keepinterval == "day") {
+        m_keepinterval = WQKI_DAY;
+    } else if (keepinterval == "week") {
+        m_keepinterval = WQKI_WEEK;
+    } else if (keepinterval == "month") {
+        m_keepinterval = WQKI_MONTH;
+    } else if (keepinterval == "year") {
+        m_keepinterval = WQKI_YEAR;
+    } else if (!keepinterval.empty()) {
+        LOGERR("WebQueueIndexer: bad value for keepinterval: " << keepinterval << "\n");
+    }
 }
 
 WebQueueIndexer::~WebQueueIndexer()
@@ -361,10 +372,35 @@ bool WebQueueIndexer::indexFiles(list<string>& files)
     return true;
 }
 
+static std::string date_string(const char *fmt)
+{
+    time_t now = time(0);
+    struct tm tmb;
+    localtime_r(&now, &tmb);
+    char buf[200];
+    strftime(buf, sizeof(buf)-1, fmt, &tmb);
+    return buf;
+}
+static std::string yearday()
+{
+    return date_string("%Y%j");
+}
+static std::string yearweek()
+{
+    return date_string("%Y%V");
+}
+static std::string yearmonth()
+{
+    return date_string("%Y%m");
+}
+static std::string yearyear()
+{
+    return date_string("%Y");
+}
+
 FsTreeWalker::Status 
-WebQueueIndexer::processone(const string &path,
-                            const struct PathStat *stp,
-                            FsTreeWalker::CbFlag flg)
+WebQueueIndexer::processone(
+    const string &path, const struct PathStat *stp, FsTreeWalker::CbFlag flg)
 {
     if (!m_db) //??
         return FsTreeWalker::FtwError;
@@ -389,6 +425,17 @@ WebQueueIndexer::processone(const string &path,
     // Have to use the hit type for the udi, because the same url can exist
     // as a bookmark or a page.
     udipath = path_cat(dotdoc.meta[Rcl::Doc::keybght], url_gpath(dotdoc.url));
+    // !! is an arbitrary separator rather unlikely to be found in urls.
+    switch (m_keepinterval) {
+    case WQKI_DAY: udipath  = udipath + "!!" + yearday(); break;
+    case WQKI_WEEK: udipath = udipath + "!!" + yearweek(); break;
+    case WQKI_MONTH: udipath= udipath + "!!" + yearmonth(); break;
+    case WQKI_YEAR: udipath = udipath + "!!" + yearyear(); break;
+    default: break;
+    }
+
+    // Also append the current date (year+day): we store one page copy per day
+    std::cerr << "UDI: " << udipath << "\n";
     make_udi(udipath, cstr_null, udi);
 
     LOGDEB("WebQueueIndexer: prc1: udi [" << udi << "]\n");

src/index/webqueue.h

@@ -65,13 +65,19 @@ public:
     bool getFromCache(const std::string& udi, Rcl::Doc &doc, std::string& data,
                       std::string *hittype = 0);
 private:
-    RclConfig *m_config;
-    Rcl::Db   *m_db;
-    WebStore  *m_cache;
+    RclConfig *m_config{nullptr};
+    Rcl::Db   *m_db{nullptr};
+    WebStore  *m_cache{nullptr};
     std::string     m_queuedir;
-    DbIxStatusUpdater *m_updater;
-    bool       m_nocacheindex;
-
+    DbIxStatusUpdater *m_updater{nullptr};
+    // Don't process the cache. Set by indexFiles().
+    bool       m_nocacheindex{false};
+    // Config: page erase interval. We normally keep only one
+    // instance. This can be set to "day", "week", "month", "year" to
+    // keep more.
+    enum KeepInterval {WQKI_NONE, WQKI_DAY, WQKI_WEEK, WQKI_MONTH, WQKI_YEAR};
+    KeepInterval  m_keepinterval{WQKI_NONE};
+    
     bool indexFromCache(const std::string& udi);
     void updstatus(const std::string& udi);
 };

src/qtgui/confgui/confguiindex.cpp

@@ -347,8 +347,7 @@ bool ConfIndexW::setupWebHistoryPanel(int idx)
     m_w->enableLink(bparam, cparam);
 
     cparam = m_w->addParam(
-        idx, ConfTabsW::CFPT_INT, "webcachemaxmbs",
-        tr("Max. size for the web store (MB)"),
+        idx, ConfTabsW::CFPT_INT, "webcachemaxmbs", tr("Max. size for the web store (MB)"),
         tr("Entries will be recycled once the size is reached."
            "<br>"
            "Only increasing the size really makes sense because "
@@ -356,6 +355,18 @@ bool ConfIndexW::setupWebHistoryPanel(int idx)
            "file (only waste space at the end)."
             ), -1, 1000*1000); // Max 1TB...
     m_w->enableLink(bparam, cparam);
+
+    QStringList intervals{"", "day", "week", "month", "year"};
+    cparam = m_w->addParam(
+        idx, ConfTabsW::CFPT_CSTR, "webcachekeepinterval", tr("Page recycle interval"),
+        tr("<p>By default, only one instance of an URL is kept in the cache. This "
+           "can be changed by setting this to a value determining at what frequency "
+           "we keep multiple instances ('day', 'week', 'month', 'year'). "
+           "Note that increasing the interval will not erase existing entries."),
+        0, 0, &intervals);
+    m_w->enableLink(bparam, cparam);
+
+
     int64_t sz = -1;
     auto ws = std::unique_ptr<WebStore>(new WebStore(m_rclconf));
     sz = ws->cc()->size();

src/sampleconf/recoll.conf

@@ -545,6 +545,16 @@ webcachemaxmbs = 40
 # then moved by a script to webqueuedir.</descr></var>
 #webdownloadsdir = ~/Downloads
 
+# <var name="webcachekeepinterval" type="string">
+#
+# <brief>Page recycle interval</brief>
+# <descr>By default, only one instance of an URL is kept in the cache. This
+# can be changed by setting this to a value determining at what frequency
+# we keep multiple instances ('day', 'week', 'month',
+# 'year'). Note that increasing the interval will not erase existing
+# entries.</descr></var>
+#webcachekeepinterval=
+
 # <var name="aspellDicDir" type="dfn">
 #
 # <brief>Aspell dictionary storage directory location.</brief> <descr>The