Add readme

.gitignore

@@ -1,2 +1,2 @@
-main
+kifflom
 big*.json

README.md

@@ -0,0 +1,51 @@
+# Kifflom!
+
+Kifflom is a (streaming) JSON parser that does not build any structures from the document,
+but returns specific values matched by given selectors instead.
+
+## Example
+
+Lets take this JSON description of fruits.
+
+```json
+{
+    "prices": {
+        "apple": 25,
+        "banana": 10,
+        "peach": 40.5,
+        "pomelo": null
+    },
+    "bananas": [
+        {"length": 13, "weight": 5},
+        {"length": 18, "weight": 8},!
+        {"length": 13, "weight": 4}
+    ]
+}
+```
+
+In order to get the weight of the first banana and prices for all the fruits
+we can use such a command:
+
+```bash
+cat test.json | kifflom -s ".bananas#0.weight .prices.*"
+# .prices.* 25
+# .prices.* 10
+# .prices.* 40.5
+# .prices.* <nil>
+#
+# Parse error! Yay!
+# [010:036] (Error: Unexpected symbol: '!')
+# .bananas#0.weight 5
+```
+
+## Performance
+
+As you can learn from benchmarks described below, kifflom's lexer itself is
+roughly 8.5 times slower than the standard JSON parser on any amount of data.
+You can benefit from low and constant memory usage, although I don't think you would.
+
+```bash
+# Running lexer tests and benchmarks
+cd lexer/
+go test -bench .
+```