summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorJohn MacFarlane <fiddlosopher@gmail.com>2012-07-25 12:31:16 -0700
committerJohn MacFarlane <fiddlosopher@gmail.com>2012-07-25 12:31:16 -0700
commit335cd5de4d785f09ae1c14e961df8c4137300252 (patch)
tree52810f2a87fb89571ad2036cc334bcda4c0c9f6e
parent0d4424c21c7d3d6da068a77750db72f61ff0e0cd (diff)
Moved stateTabStop to readerTabStop in ReaderOptions.
-rw-r--r--src/Text/Pandoc/Options.hs2
-rw-r--r--src/Text/Pandoc/Parsing.hs3
-rw-r--r--src/Text/Pandoc/Readers/Markdown.hs13
-rw-r--r--src/Text/Pandoc/Readers/RST.hs4
-rw-r--r--src/pandoc.hs4
5 files changed, 11 insertions, 15 deletions
diff --git a/src/Text/Pandoc/Options.hs b/src/Text/Pandoc/Options.hs
index 28a962607..ab4dda388 100644
--- a/src/Text/Pandoc/Options.hs
+++ b/src/Text/Pandoc/Options.hs
@@ -56,6 +56,7 @@ data ReaderOptions = ReaderOptions{
, readerStrict :: Bool -- ^ FOR TRANSITION ONLY
, readerParseRaw :: Bool -- ^ Parse raw HTML, LaTeX
, readerColumns :: Int -- ^ Number of columns in terminal
+ , readerTabStop :: Int -- ^ Tab stop
} deriving (Show, Read)
instance Default ReaderOptions
@@ -65,4 +66,5 @@ instance Default ReaderOptions
, readerStrict = False
, readerParseRaw = False
, readerColumns = 80
+ , readerTabStop = 4
}
diff --git a/src/Text/Pandoc/Parsing.hs b/src/Text/Pandoc/Parsing.hs
index 2d7c4eb81..3d7057936 100644
--- a/src/Text/Pandoc/Parsing.hs
+++ b/src/Text/Pandoc/Parsing.hs
@@ -544,7 +544,6 @@ tableWith headerParser rowParser lineParser footerParser = try $ do
(heads, aligns, indices) <- headerParser
lines' <- rowParser indices `sepEndBy1` lineParser
footerParser
- state <- getState
numColumns <- getOption readerColumns
let widths = if (indices == [])
then replicate (length aligns) 0.0
@@ -696,7 +695,6 @@ data ParserState = ParserState
stateKeys :: KeyTable, -- ^ List of reference keys
stateCitations :: [String], -- ^ List of available citations
stateNotes :: NoteTable, -- ^ List of notes
- stateTabStop :: Int, -- ^ Tab stop
stateStandalone :: Bool, -- ^ Parse bibliographic info?
stateTitle :: [Inline], -- ^ Title of document
stateAuthors :: [[Inline]], -- ^ Authors of document
@@ -729,7 +727,6 @@ defaultParserState =
stateKeys = M.empty,
stateCitations = [],
stateNotes = [],
- stateTabStop = 4,
stateStandalone = False,
stateTitle = [],
stateAuthors = [],
diff --git a/src/Text/Pandoc/Readers/Markdown.hs b/src/Text/Pandoc/Readers/Markdown.hs
index 69faadd4a..4f1831baf 100644
--- a/src/Text/Pandoc/Readers/Markdown.hs
+++ b/src/Text/Pandoc/Readers/Markdown.hs
@@ -85,15 +85,13 @@ isBlank _ = False
indentSpaces :: Parser [Char] ParserState [Char]
indentSpaces = try $ do
- state <- getState
- let tabStop = stateTabStop state
+ tabStop <- getOption readerTabStop
count tabStop (char ' ') <|>
string "\t" <?> "indentation"
nonindentSpaces :: Parser [Char] ParserState [Char]
nonindentSpaces = do
- state <- getState
- let tabStop = stateTabStop state
+ tabStop <- getOption readerTabStop
sps <- many (char ' ')
if length sps < tabStop
then return sps
@@ -101,8 +99,8 @@ nonindentSpaces = do
skipNonindentSpaces :: Parser [Char] ParserState ()
skipNonindentSpaces = do
- state <- getState
- atMostSpaces (stateTabStop state - 1)
+ tabStop <- getOption readerTabStop
+ atMostSpaces (tabStop - 1)
atMostSpaces :: Int -> Parser [Char] ParserState ()
atMostSpaces 0 = notFollowedBy (char ' ')
@@ -627,8 +625,7 @@ defListMarker :: Parser [Char] ParserState ()
defListMarker = do
sps <- nonindentSpaces
char ':' <|> char '~'
- st <- getState
- let tabStop = stateTabStop st
+ tabStop <- getOption readerTabStop
let remaining = tabStop - (length sps + 1)
if remaining > 0
then count remaining (char ' ') <|> string "\t"
diff --git a/src/Text/Pandoc/Readers/RST.hs b/src/Text/Pandoc/Readers/RST.hs
index 675524443..b017b736b 100644
--- a/src/Text/Pandoc/Readers/RST.hs
+++ b/src/Text/Pandoc/Readers/RST.hs
@@ -33,6 +33,7 @@ module Text.Pandoc.Readers.RST (
import Text.Pandoc.Definition
import Text.Pandoc.Shared
import Text.Pandoc.Parsing
+import Text.Pandoc.Options
import Control.Monad ( when, liftM, guard, mzero )
import Data.List ( findIndex, intercalate, transpose, sort, deleteFirstsBy )
import qualified Data.Map as M
@@ -493,8 +494,7 @@ listLine markerLength = try $ do
-- indent by specified number of spaces (or equiv. tabs)
indentWith :: Int -> Parser [Char] ParserState [Char]
indentWith num = do
- state <- getState
- let tabStop = stateTabStop state
+ tabStop <- getOption readerTabStop
if (num < tabStop)
then count num (char ' ')
else choice [ try (count num (char ' ')),
diff --git a/src/pandoc.hs b/src/pandoc.hs
index fe7b54f9d..a3c01133c 100644
--- a/src/pandoc.hs
+++ b/src/pandoc.hs
@@ -936,8 +936,7 @@ main = do
then "."
else takeDirectory (head sources)
- let startParserState = def{ stateTabStop = tabStop,
- stateLiterateHaskell = "+lhs" `isSuffixOf` readerName' ||
+ let startParserState = def{ stateLiterateHaskell = "+lhs" `isSuffixOf` readerName' ||
lhsExtension sources,
stateStandalone = standalone',
stateCitations = map CSL.refId refs,
@@ -948,6 +947,7 @@ main = do
(laTeXOutput || writerName' == "context"))
, readerParseRaw = parseRaw
, readerColumns = columns
+ , readerTabStop = tabStop
},
stateIndentedCodeClasses = codeBlockClasses,
stateApplyMacros = not laTeXOutput