summaryrefslogtreecommitdiff
path: root/src/io/extract.cpp
diff options
context:
space:
mode:
authorBen Longbons <b.r.longbons@gmail.com>2014-10-25 15:24:26 -0700
committerBen Longbons <b.r.longbons@gmail.com>2014-10-26 14:21:48 -0700
commit86395f53634b3ef1ce76a7f1e5edfdb61f8ffd80 (patch)
tree2710c62fe71d5e0d2e228fba9c951a040c4dcddf /src/io/extract.cpp
parent6800761863dd45b6055768febc6ace6a20120dc7 (diff)
downloadtmwa-86395f53634b3ef1ce76a7f1e5edfdb61f8ffd80.tar.gz
tmwa-86395f53634b3ef1ce76a7f1e5edfdb61f8ffd80.tar.bz2
tmwa-86395f53634b3ef1ce76a7f1e5edfdb61f8ffd80.tar.xz
tmwa-86395f53634b3ef1ce76a7f1e5edfdb61f8ffd80.zip
Fix header ranking
Diffstat (limited to 'src/io/extract.cpp')
-rw-r--r--src/io/extract.cpp220
1 files changed, 220 insertions, 0 deletions
diff --git a/src/io/extract.cpp b/src/io/extract.cpp
new file mode 100644
index 0000000..55d3980
--- /dev/null
+++ b/src/io/extract.cpp
@@ -0,0 +1,220 @@
+#include "extract.hpp"
+// extract.cpp - a simple, hierarchical, tokenizer
+//
+// Copyright © 2013-2014 Ben Longbons <b.r.longbons@gmail.com>
+//
+// This file is part of The Mana World (Athena server)
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program. If not, see <http://www.gnu.org/licenses/>.
+
+#include <algorithm>
+
+#include "../strings/astring.hpp"
+#include "../strings/xstring.hpp"
+#include "../strings/vstring.hpp"
+
+#include "../poison.hpp"
+
+
+// TODO also pass an io::LineSpan around.
+namespace tmwa
+{
+bool extract(XString str, XString *rv)
+{
+ *rv = str;
+ return true;
+}
+
+bool extract(XString str, RString *rv)
+{
+ *rv = str;
+ return true;
+}
+
+bool extract(XString str, AString *rv)
+{
+ *rv = str;
+ return true;
+}
+
+bool extract(XString str, std::chrono::nanoseconds *ns)
+{
+ std::chrono::nanoseconds::rep rep;
+ if (extract(str, &rep))
+ {
+ *ns = std::chrono::nanoseconds(rep);
+ return true;
+ }
+ if (str.endswith("ns"_s))
+ {
+ if (extract(str.xrslice_h("ns"_s.size()), &rep))
+ {
+ *ns = std::chrono::nanoseconds(rep);
+ return true;
+ }
+ return false;
+ }
+ std::chrono::microseconds bigger;
+ if (extract(str, &bigger))
+ {
+ *ns = bigger;
+ return *ns == bigger;
+ }
+ return false;
+}
+bool extract(XString str, std::chrono::microseconds *us)
+{
+ std::chrono::microseconds::rep rep;
+ if (extract(str, &rep))
+ {
+ *us = std::chrono::microseconds(rep);
+ return true;
+ }
+ if (str.endswith("us"_s))
+ {
+ if (extract(str.xrslice_h("us"_s.size()), &rep))
+ {
+ *us = std::chrono::microseconds(rep);
+ return true;
+ }
+ return false;
+ }
+ std::chrono::milliseconds bigger;
+ if (extract(str, &bigger))
+ {
+ *us = bigger;
+ return *us == bigger;
+ }
+ return false;
+}
+bool extract(XString str, std::chrono::milliseconds *ms)
+{
+ std::chrono::milliseconds::rep rep;
+ if (extract(str, &rep))
+ {
+ *ms = std::chrono::milliseconds(rep);
+ return true;
+ }
+ if (str.endswith("ms"_s))
+ {
+ if (extract(str.xrslice_h("ms"_s.size()), &rep))
+ {
+ *ms = std::chrono::milliseconds(rep);
+ return true;
+ }
+ return false;
+ }
+ std::chrono::seconds bigger;
+ if (extract(str, &bigger))
+ {
+ *ms = bigger;
+ return *ms == bigger;
+ }
+ return false;
+}
+bool extract(XString str, std::chrono::seconds *s)
+{
+ std::chrono::seconds::rep rep;
+ if (extract(str, &rep))
+ {
+ *s = std::chrono::seconds(rep);
+ return true;
+ }
+ if (str.endswith("s"_s))
+ {
+ if (extract(str.xrslice_h("s"_s.size()), &rep))
+ {
+ *s = std::chrono::seconds(rep);
+ return true;
+ }
+ return false;
+ }
+ std::chrono::minutes bigger;
+ if (extract(str, &bigger))
+ {
+ *s = bigger;
+ return *s == bigger;
+ }
+ return false;
+}
+bool extract(XString str, std::chrono::minutes *min)
+{
+ std::chrono::minutes::rep rep;
+ if (extract(str, &rep))
+ {
+ *min = std::chrono::minutes(rep);
+ return true;
+ }
+ if (str.endswith("min"_s))
+ {
+ if (extract(str.xrslice_h("min"_s.size()), &rep))
+ {
+ *min = std::chrono::minutes(rep);
+ return true;
+ }
+ return false;
+ }
+ std::chrono::hours bigger;
+ if (extract(str, &bigger))
+ {
+ *min = bigger;
+ return *min == bigger;
+ }
+ return false;
+}
+bool extract(XString str, std::chrono::hours *h)
+{
+ std::chrono::hours::rep rep;
+ if (extract(str, &rep))
+ {
+ *h = std::chrono::hours(rep);
+ return true;
+ }
+ if (str.endswith("h"_s))
+ {
+ if (extract(str.xrslice_h("h"_s.size()), &rep))
+ {
+ *h = std::chrono::hours(rep);
+ return true;
+ }
+ return false;
+ }
+ std::chrono::duration<int, std::ratio<60*60*24>> bigger;
+ if (extract(str, &bigger))
+ {
+ *h = bigger;
+ return *h == bigger;
+ }
+ return false;
+}
+bool extract(XString str, std::chrono::duration<int, std::ratio<60*60*24>> *d)
+{
+ std::chrono::duration<int, std::ratio<60*60*24>>::rep rep;
+ if (extract(str, &rep))
+ {
+ *d = std::chrono::duration<int, std::ratio<60*60*24>>(rep);
+ return true;
+ }
+ if (str.endswith("d"_s))
+ {
+ if (extract(str.xrslice_h("d"_s.size()), &rep))
+ {
+ *d = std::chrono::duration<int, std::ratio<60*60*24>>(rep);
+ return true;
+ }
+ return false;
+ }
+ return false;
+}
+} // namespace tmwa