1 files changed, 109 insertions, 0 deletions
diff --git a/cepis/csv.lua b/cepis/csv.lua
new file mode 100644
index 0000000..be05c9a
--- /dev/null
+++ b/cepis/csv.lua
@@ -0,0 +1,109 @@
+--- A minimal RFC 4180 reader, for the batch endpoint.
+--
+-- The values the API sends may contain commas — a `logradouro` such as
+-- `Rua Um, Dois` — and are then double-quoted with `""` escaping. Splitting on
+-- `,` silently misaligns every column after the first such value, which is why
+-- this exists instead of a `gmatch`.
+--
+-- @module cepis.csv
+
+local M = {}
+
+local BOM = "\239\187\191"
+
+--- Parse CSV text into a list of rows, each a list of field strings.
+--
+-- Handles quoted fields, `""` escapes, embedded commas and newlines, both LF
+-- and CRLF line endings, and a leading UTF-8 BOM. Blank lines yield no row.
+-- Nothing is coerced: every field comes back as a string.
+--
+-- Text between a closing quote and the next delimiter is appended to the field
+-- rather than starting a new one: malformed input must not silently gain a
+-- column and shift every value after it.
+--
+-- @tparam string text the response body
+-- @treturn table list of rows; each row is a list of strings
+function M.parse(text)
+  if text:sub(1, 3) == BOM then
+    text = text:sub(4)
+  end
+
+  local rows = {}
+  local row = {}
+  local count = 0
+  local pos = 1
+  local size = #text
+
+  local function close_row()
+    if not (count == 1 and row[1] == "") then
+      rows[#rows + 1] = row
+    end
+    row = {}
+    count = 0
+  end
+
+  while pos <= size do
+    local field
+    if text:sub(pos, pos) == '"' then
+      local parts = {}
+      pos = pos + 1
+      while true do
+        local quote = text:find('"', pos, true)
+        if not quote then
+          parts[#parts + 1] = text:sub(pos)
+          pos = size + 1
+          break
+        end
+        parts[#parts + 1] = text:sub(pos, quote - 1)
+        if text:sub(quote + 1, quote + 1) == '"' then
+          parts[#parts + 1] = '"'
+          pos = quote + 2
+        else
+          pos = quote + 1
+          break
+        end
+      end
+      local trailing = text:find("[,\r\n]", pos)
+      if trailing ~= pos then
+        parts[#parts + 1] = text:sub(pos, (trailing or size + 1) - 1)
+        pos = trailing or size + 1
+      end
+      field = table.concat(parts)
+    else
+      local stop = text:find("[,\r\n]", pos)
+      if stop then
+        field = text:sub(pos, stop - 1)
+        pos = stop
+      else
+        field = text:sub(pos)
+        pos = size + 1
+      end
+    end
+
+    count = count + 1
+    row[count] = field
+
+    local char = text:sub(pos, pos)
+    if char == "," then
+      pos = pos + 1
+      if pos > size then
+        count = count + 1
+        row[count] = ""
+        close_row()
+      end
+    elseif char == "\r" or char == "\n" then
+      if char == "\r" and text:sub(pos + 1, pos + 1) == "\n" then
+        pos = pos + 2
+      else
+        pos = pos + 1
+      end
+      close_row()
+    else
+      close_row()
+    end
+  end
+
+  return rows
+end
+
+return M