aboutsummaryrefslogtreecommitdiff
path: root/src/Text/Pandoc
diff options
context:
space:
mode:
authorJohn MacFarlane <jgm@berkeley.edu>2015-10-12 22:04:40 -0700
committerJohn MacFarlane <jgm@berkeley.edu>2015-10-12 22:04:40 -0700
commit476b383c578699567ac4630391a15855521ab3d4 (patch)
tree67c25848255ecde5ff835b630532e059d36daf3d /src/Text/Pandoc
parent0715c5cbedf926fc2b3546b8f964af0b4b74e9ee (diff)
downloadpandoc-476b383c578699567ac4630391a15855521ab3d4.tar.gz
RST writer: tweaks to header normalization.
These changes are intended to make the writer more useful to people who are processing small fragments, which may for example look like this: ### third level header from previous section ## second level header Previously such fragments got turned into two headers of the same level. The new algorithm avoids doing any normalization until we hit the minimal-level header in the fragment (here, the second level header). Closes #2394.
Diffstat (limited to 'src/Text/Pandoc')
-rw-r--r--src/Text/Pandoc/Writers/RST.hs26
1 files changed, 22 insertions, 4 deletions
diff --git a/src/Text/Pandoc/Writers/RST.hs b/src/Text/Pandoc/Writers/RST.hs
index 754aee29c..70caa319f 100644
--- a/src/Text/Pandoc/Writers/RST.hs
+++ b/src/Text/Pandoc/Writers/RST.hs
@@ -82,7 +82,7 @@ pandocToRST (Pandoc meta blocks) = do
(fmap (trimr . render colwidth) . inlineListToRST)
$ deleteMeta "title" $ deleteMeta "subtitle" meta
let minLev = findMinHeadingLevel Nothing blocks
- body <- blockListToRST' True $ normalizeHeadings minLev blocks
+ body <- blockListToRST' True $ normalizeHeadings minLev 0 blocks
notes <- liftM (reverse . stNotes) get >>= notesToRST
-- note that the notes may contain refs, so we do them first
refs <- liftM (reverse . stLinks) get >>= refsToRST
@@ -102,12 +102,30 @@ pandocToRST (Pandoc meta blocks) = do
then return $ renderTemplate' (writerTemplate opts) context
else return main
where
- normalizeHeadings lev (Header l a i:bs) = Header lev a i:normalizeHeadings (lev+1) cont ++ normalizeHeadings lev bs'
+ -- Partial normalization:
+ -- We use the underline scheme appropriate for the lowest
+ -- header level found, and we don't normalize headers
+ -- until we've hit the first header with minimum header
+ -- level in the fragment.
+ -- This is intended to help with the problems noted in
+ -- #2079, while still being flexible enough for use
+ -- with fragments that might not include the base
+ -- header level, and might start halfway through a section
+ -- (see #2394).
+ normalizeHeadings minLev lev (Header l a i:bs) =
+ let lev' = if lev == 0 && l == minLev
+ then minLev
+ else lev
+ in if lev' == 0 -- we haven't hit minLev yet
+ then Header l a i : normalizeHeadings minLev 0 cont ++
+ normalizeHeadings minLev 0 bs'
+ else Header lev' a i : normalizeHeadings minLev (lev' + 1) cont ++
+ normalizeHeadings minLev lev' bs'
where (cont,bs') = break (headerLtEq l) bs
headerLtEq level (Header l' _ _) = l' <= level
headerLtEq _ _ = False
- normalizeHeadings lev (b:bs) = b:normalizeHeadings lev bs
- normalizeHeadings _ [] = []
+ normalizeHeadings minLev lev (b:bs) = b : normalizeHeadings minLev lev bs
+ normalizeHeadings _ _ [] = []
findMinHeadingLevel Nothing (Header l _a _i:bs) = findMinHeadingLevel (Just l) bs
findMinHeadingLevel (Just ol) (Header l _a _i:bs) =
findMinHeadingLevel (Just $ if ol>l then l else ol) bs