Gutcheck documentation
gutcheck: lists possible common formatting errors in a Project
Gutenberg candidate file. It is a command line program and can be used
under Win32 or Unix (gutcheck.c should compile anywhere; if it doesn't,
tell me). For Windows-only people, there is an appendix at the end
with brief instructions for running it.
Current version: 0.99. Users of 0.98 see end of file for changes.
You should also have received the licence file COPYING, a README file,
gutcheck.c, the source code, and gutcheck.exe, a DOS executable, with
this file.
This software is Copyright Jim Tinsley 2000-2005.
Gutcheck comes wih ABSOLUTELY NO WARRANTY. For details, read the file COPYING.
This is Free Software; you may redistribute it under certain conditions (GPL).
See http://gutcheck.sourceforge.net for the latest version.
Usage is: gutcheck [-setopxlywm] filename
where:
-s checks Single quotes
-e switches off Echoing of lines
-t checks Typos
-o produces an Overview only
-p sets strict quotes checking for Paragraphs
-x (paranoid) switches OFF typo checking and extra checks
-l turns off Line-end checks
-y sets error messages to stdout
-w is a special mode for web uploads (for future use)
-v (verbose) forces individual reporting of minor problems
-m interprets Markup of some common HTML tags and entities
-u warns about words in a user-defined typo file gutcheck.typ
-d ignores some DP-specific markup
Running gutcheck without any parameters will display a brief help message.
Sample usage:
gutcheck warpeace.txt
More detail:
Echoing lines (-e to switch off)
You may find it convenient, when reviewing Gutcheck's
suggestions, to see the line that Gutcheck is questioning.
That way, you can often see at a glance whether it is
a real error that needs to be fixed, or a false positive
that should be in the text, but Gutcheck's limited
programming doesn't understand.
By default, gutcheck echoes these lines, but if you don't
want to see the lines referred to, -e will switch it OFF.
Quotes (-s and -p switches)
Gutcheck always looks for unbalanced doublequotes in a
paragraph. It is a common convention for writers not to
close quotes in a paragraph if the next paragraph opens
with quotes and is a continuation by the same speaker.
Gutcheck therefore does not normally report unclosed quotes
if the next paragraph begins with a quote. If you need
to see all unclosed quotes, even where the next paragraph
begins with a quote, you should use the -p switch.
Singlequotes (') are a problem, since the same character
is used for an apostrophe. I'm not sure that it is
possible to get 100% accuracy on singlequotes checking,
particularly since dialect, quite common in PG texts,
upsets the normal rules so badly. Consider the sentence:
'Tis often said that a man's a man for a' that.
As humans, we recognize that both apostrophes are used
for contractions rather than quotes, but it isn't easy
to get a program to recognize that.
Since Gutcheck makes too many mistakes when trying to match
singlequotes, it doesn't look for unbalanced singlequotes
unless you specify the -s switch.
Consider these sentences, which illustrate the main cases:
'Tis often said that a fool and his money are soon parted.
'Becky's goin' home,' said Tom.
The dogs' tails wagged in unison.
Those 'pack dogs' of yours look more like wolves.
Typos (-t switch)
It's not Gutcheck's job to be a spelling checker, but it
does check for a list of common typos and OCR errors if you
use the -t switch. (The -x switch also turns typo checking on.)
It also checks for character combinations, especially involving
h and b, which are often confused by OCR, that rarely or never
occur. For example, it queries "tbe" in a word. Now, "the" often
occurs, but "tbe" is very rare (heartbeat, hotbed), so I'm
playing the odds - a few false positives for many errors found.
Similarly with "ii", which is a very common OCR error.
Gutcheck suppresses multiple reporting of the first 40 "typos"
found. This is to remove the annoyance of seeing something like
"FN" (footnote) or "LK" (initials) flagged as a typo 147 times
in a text.
Line-end checking (-l switch to disable)
All PG texts should have a Carriage Return (CR - character 13)
and a Line Feed (LF - character 10) at end of each line,
regardless of what O/S you made them on. DOS/Windows, Unix
and Mac have different conventions, but the final text should
always use a CR/LF pair as its line terminator.
By default, Gutcheck verifies that every line does have
the correct terminator, but if you're on a work-in-progress
in Linux, you might want to convert the line-ends as a final
step, and not want to see thousands of errors every time you
run Gutcheck before that final step, so you can turn off
this checking with the -l switch.
Paranoid mode (-x switch to disable: Trust No One :-)
-x switches OFF typo-checking, the -t flag, automatically
and some extra checks like standalone 1 and 0 queries.
Overview mode (-o switch)
This mode just gives a count of queries found
instead of a detailed list.
Header quote (-h switch)
If you use the -h switch, gutcheck will also display
the Title, Author, Release and Edition fields from the
PG header. This is useful mostly for the automated
checks we do on recently-posted texts.
Errors to stdout (-y switch)
If you're just running gutcheck normally, you can ignore
this. It's only there for programs that provide a front
end to gutcheck. It makes error messages appear within
the output of gutcheck so that the front end knows whether
gutcheck ran OK.
Verbose reporting (-v switch)
Normally, if gutcheck sees lots of long lines, short lines,
spaced dashes, non-ASCII characters or dot-commas ".," it
assumes these are features of the text, counts and summarizes
them at the top of its report, but does not list them
individually. If the -v switch is on, gutcheck will list them all.
Markup interpretation (-m switch)
Normally, gutcheck flags anything it suspects of being HTML
markup as a possible error. When you use the -m switch,
however, it matches anything that looks like markup against
a short list of common HTML tags and entities. If the markup
is in that list, it either ignores the markup, in the case
of a tag, or "interprets" the markup as its nearest ASCII
equivalent, in the case of an entity. So, for example, using
this switch, gutcheck will "see"
“He went thataway!”
as
"He went thataway!"
and report accordingly.
This switch does not, not, NOT check the validity of HTML;
it exists so that you can run gutcheck on most HTML texts
for PG, and get sane results. It does not support all tags.
It does not support all entities. When it sees a tag or entity
it does not recognize, it will query it as HTML just as if
you hadn't specified the -m switch.
Gutcheck 0.99 will automatically switch on markup interpretation
if it sees a lot of tags that appear to be markup, so mostly, you
won't have to specify this.
User-defined typos (-u switch)
If you have a file named gutcheck.typ either in your current
working directory or in the directory from which you explicitly
invoked gutcheck, but not necessarily on your path, and if you
specify the -u switch, gutcheck will query any word specified
in that file. The file is simple: one word, in lower case, per
line. 999 lines are allowed for. Be careful not to put multiple
words onto a line, or leave any rubbish other than the word on
the line. You should have received a sample file gutcheck.typ
with this package.
Ignore DP markup (-d switch)
Distributed Proofreaders (http://www.pgdp.net) is currently
(2005) the main source of PG texts, and proofers there use
special conventions. This switch understands those conventions,
so that people can use gutcheck on files in process that still
haven't had the special conventions removed yet. The special
conventions supported in 0.99 are page-separators and
"", "", "/*", "*/", "/#", "#/", "/$", "$/".
You will probably only run gutcheck on a text once or maybe twice,
just prior to uploading; it usually finds a few formatting problems;
it also usually finds queries that aren't problems at all - it often
questions Tables of Contents for having short lines, for example.
These are called "false positives", and need a human to decide on
them.
The text should be standard prose, and already close to PG normal
format (plain text, about 70 characters per line with blank lines
between paragraphs).
Gutcheck merely draws your attention to things that might be errors.
It is NOT a substitute for human judgement. Formatting choices like
short lines may be for a reason that this program can't understand.
Even the most careful human proofing can leave errors behind in a
text, and there are several automated checks you can do to help find
them. Of these, spellchecking (with _very_ careful human judgement) is
the most important and most useful.
Gutcheck does perform some basic typo-checking if you ask it to,
but its focus is on formatting errors specific to PG texts -
mismatched quotes, non-ASCII characters, bad spacing, bad line
length, HTML tags perhaps left from a conversion, unbalanced
brackets.
Suggestions for additional checks would be appreciated and duly
considered, but no guarantees that they will be implemented.
How do _I_ use it?
Practically everyone I give gutcheck to asks me how _I_ use it.
Well, when I get a text for posting, say filename.txt, I run
gutcheck -o filename.txt
That gives me a quick idea what I'm dealing with. It'll tell
me what kind of problems gutcheck sees, and give me an idea
of how much more work needs to be done on the text. Keep in
mind that gutcheck doesn't do anything like a full spellcheck,
but when I see a text that has a lot of problems, I assume that
it probably needs a spellcheck too.
Having got a feel for the ballpark, I run
gutcheck filename.txt > jj
where jj is my personal, all-purpose filename for temporary data
that doesn't need to be kept. Then I open filename.txt and jj in
a split-screen view in my editor, and work down the text, fixing
whatever needs fixing, and skipping whatever doesn't. If your
editor doesn't split-screen, you can get much the same effect by
opening your original file in your normal editor, and jj (or your
equivalent name) in something like Notepad, keeping both in view
at the same time.
Twice a day, an automatic process looks at all recently-posted
texts, and emails Michael, me, and sometimes other people with
their gutcheck summaries.
Future development of gutcheck
Gutcheck has gone about as far as it can, given its current
structure. In order to add better singlequotes checking,
sentence checking, better he/be checking and other good stuff
that I'd like to see, I'll have to rewrite it from a different
angle - looking at the syntax instead of the lines. And I'll
probably get around to that sooner or later.
Meantime, I'm just trying to get this version stabilized, so
please report any bugs you find. When it is stable, I'll run
up a Windows port for those timid souls who can't look a
command line in the eye. :-)
And I've started work on gutspell, a companion to gutcheck
which will concentrate on spelling problems. PG spelling
problems are unusual, since the range of texts we cover is
so wide, and I'll be taking a somewhat unorthodox approach
to writing this spelling-checker _specifically_ for texts
containing a lot of dialect and uncommon words that have
probably already been spell-checked against a standard
modern dictionary.
Explanations of common gutcheck messages:
--> 74 lines in this file have white space at end
PG texts shouldn't have extra white space added at end of line.
Don't worry too much about this; they're not doing any harm,
and they'll be removed during posting anyway.
--> 348 lines in this file are short. Not reporting short lines.
--> 84 lines in this file are long. Not reporting long lines.
--> 8 lines in this file are VERY long!
If there are a lot of long or short lines, Gutcheck won't list
them individually. The short lines version of this message
is commonly seen when gutchecking poetry and some plays, where
the normal line length is shorter than the standard for prose.
A "VERY long" line is one over 80 characters. You normally
shouldn't have any of these, but sometimes you may have to render
a table that must be that long, or some special preformatted
quotation that can't be broken.
--> There are 75 spaced dashes and em-dashes in this file. Not reporting them.
The PG standard for an emdash--like these--is two minus signs
with no spaces before or after them. However, some older texts
used spaced dashes - like these -- and if there are very many
such spaced dashes in the file, gutcheck just draws your
attention to it and doesn't list them individually.
Line 3020 - Non-ASCII character 233
Standard PG texts should use only ASCII characters with values
up to 127; however, non-English, accented characters can be
represented according to several different non-ASCII encoding
schemes, using values over 127. If you have a plain English text
with a few accented characters in words like cafe or tete-a-tete,
you should replace the accented characters with their unaccented
versions. The English pound sign is another commonly-seen
non-ASCII character. If you have enough non-ASCII characters in
your text that you feel removing them would degrade your text
unacceptably, you should probably consider doing an 8-bit text
as well as a plain-ASCII version.
Line 1207 - Non-ISO-8859 character 156
Even in "8-bit" texts, there are distinctions between code sets.
The ISO-8859 family of 8-bit code sets is the most commonly used
in PG, and these sets do not define values in the range 128 through
159 as printable characters. It's quite common for someone on a
Windows or Mac machine to use a non-ISO character inadvertently,
so this message warns that the character is not only not ASCII,
but also outside the ISO-8859 range.
Line 46 - Tab character?
Some editors and WPs will put in Tab characters (character 9) to
indicate indented text. You should not use these in a PG text,
because you can't be sure how they will appear on a reader's
screen. Find the Tab, and replace it with the appropriate number
of spaces.
Line 1327 - Tilde character?
The tilde character (~) might be legitimately used, but it's the
character commonly used by OCR software to indicate a place where
it couldn't make out the letter, so gutcheck flags it.
Line 1347 - Asterisk?
Asterisks are reported only in paranoid mode (see -x).
Like tildes, they are often used to indicate errors, but they are
also legitimately used as line delimiters and footnote markers.
Line 1451 - Long line 129
PG texts should have lines shorter than 76. There may be occasions
where you decide that you really have to go out to 79 characters,
but the sample above says that line 1451 is 129 characters long -
probably two lines run together.
Line 1590 - Short line?
PG texts should have lines longer than 54 characters. However,
there are special cases like poetry and tables of contents where
the lines _should_ be shorter. So treat Gutcheck warnings about
short lines carefully. Sometimes it's a genuine formatting
problem; sometimes the line really needs to be short.
Hint: gutcheck will not flag lines as short if they are indented
- if they start with a space. I like to start inserted stanzas
and other such items indented with a couple of spaces so that
they stand out from the main text anyway.
Line 1804 - Begins with punctuation?
Lines should normally not begin with commas, periods and so on.
An exception is ellipses . . . which can happen at start of line.
Line 1850 - Spaced em-dash?
The PG standard for an em-dash--like these--is two minus signs
with no spaces before or after them. Gutcheck flags non-PG
em-dashes - like this one. Normally, you will replace it with a
PG-standard em-dash.
Line 1904 - Query he/be error?
Gutcheck makes a very minor effort to look for that scourge of all
proofreaders, "be" replacing "he" or vice-versa, and draws your
attention to it when it thinks it has found one.
Line 2017 - Query digit in a1most
The digit 1 is commonly OCRed for the letter l, the digit 0 for
the letter O, and so on. When gutcheck sees a mix of digits and
letters, it warns you. It may generate a false positive for
something like 7am.
Line 2083 - Query standalone 0
In paranoid mode (see -x) only, gutcheck warns about the digit 0
and the number 1 standing alone as a word. This can happen if the
OCR misreads the words O or I.
Line 2115 - Query word whetber
If you have switched typo-checking on, gutcheck looks for
potential typos, especially common h/b errors. It's not
infallible; it sometimes queries legit words, but it's
always worth taking a look.
Line 2190 column 14 - Missing space?
Omitting a space is a very common error,especially coming from
OCRed text,and can be hard for a human to spot. The commas in
the previous sentence illustrate the kind of thing I mean.
Line 2240 column 48 - Spaced punctuation?
The flip side of the "missing space" error , here , is when extra
spaces are added before punctuation . Some old texts appear to add
extra spaces around punctuation consistently, but this was a
typographical convention rather than the author's intent, and the
extra "spaces" should be removed when preparing a PG text.
Line 2301 column 19 - Unspaced quotes?
Another common spacing problem occurs in a phrase like "You wait
there,"he said.
Line 2385 column 27 - Wrongspaced quotes?
As of version 0.98, gutcheck adds extra checks on whether a quote
seems to be a start or end quote, and queries those that appear to
be misplaced. This does give rise to false positives when quotes are
nested, for example:
"And how," she asked, "will your "friends" help you now?"
but these false positives are worth it because of the many cases
that this test catches, notably those like:
"And how, "she said," will your friends help you now?"
Sometimes a "wrongspaced quotes" query will arise because an earlier
quote in the paragraph was omitted, so if the place specified seems
to be OK, look back to see whether there's a problem in the preceding
lines.
Line 2400 - HTML Tag?
Some PG texts have been converted from HTML, and not all of the
HTML tags have been removed.
Line 2402 - HTML symbol? &emdash;
Similarly, special HTML symbol characters can survive into PG
texts. Can occasionally produce amusing false positives like
. . . Marwick & Co were well known for it;
Line 2540 - Mismatched quotes
Another gutcheck mainstay - unclosed doublequotes in a paragraph.
See the discussion of quotes in the switches section near the
start of this file.
Since the mismatch doesn't occur on any one line, gutcheck quotes
the line number of the first blank line following the paragraph,
since this is the point where it reconciles the count of quotes.
However, if gutcheck is echoing lines, that is, you haven't used
the -e switch, it will show the _first_ line of the paragraph,
to help you find the place without using line numbers. The
offending paragraph is therefore between the quoted line and
the line number given.
Line 2587 - Mismatched single quotes
Only checked with the -s switch, since checking single quotes is
not a very reliable process. Otherwise, the same logic as for
doublequotes applies.
Line 2877 - Mismatched round brackets?
Also curly and square brackets. Texts with a lot of brackets, like
plays with bracketed stage instructions, may have mismatches.
Line 3150 - No CR?
Line 3204 - Two successive CRs?
Line 3281 position 75 - CR without LF?
These are the invalid line-end warnings. See the discussion of
line-end checking in the switches section near the start of this
file. If you see these, and your editor doesn't show anything
wrong, you should probably try deleting the characters just before
and after the line end, and the line-end itself, then retyping the
characters and the line-end.
Line 2940 - Paragraph starts with lower-case
A common error in an e-text is for an extra blank line
to be put in, like the blank line above, and this often
shows up as a new paragraph beginning with lower case.
Sometimes the blank line is deliberate, as when a
quotation is inserted in a speech. Use your judgement.
Line 2987 - Extra period?
An extra period. is a. common problem in OCRed text. and usually
arises when a speck of dust on the page is mistaken for a period.
or. as occasionally happens. when a comma loses its tail.
Line 3012 column 12 - Double punctuation?
Double punctuation., like that,, is a common typo and
scanno. Some books have much legit double punctuation,
like etc., etc., but it's worth checking anyway.
* * * *
For Windows-only users who are unfamiliar with DOS:
If you're a Windows-only user, you need to save
gutcheck.exe into the folder (directory) where the
text file you want to check is. Let's say your
text file is in C:\GUT, then you should save
GUTCHECK.EXE into C:\GUT.
Now get to a DOS prompt. You can do this by
selecting the "Command Prompt" or "MS-DOS Prompt"
option that will be somewhere on your
Start/Programs menu.
Now get into the C:\GUT directory.
You can do this using the CD (change directory)
command, like this:
CD \GUT
and your prompt will change to
C:\GUT>
so you know you're in the right place.
Now type
gutcheck yourfile.txt
and you'll see gutcheck's report
By default, gutcheck prints its queries to screen.
If you want to create a file of them, to edit
against the text, you can use the greater-than
sign (>) to tell it to output the report to a
file. For example, if you want its report in a
file called QUERIES.LST, you could type
gutcheck yourfile.txt > queries.lst
The queries.lst file will then contain the listing
of possible formatting errors, and you can
edit it alongside your text.
Whatever you do, DON'T make the filename after
the greater-than sign the name of a file already
on your disk that you want to keep, because
the greater-than sign will cause gutcheck to
replace any existing file of that name.
So, for example, if you have two Tolstoy files
that you want to check, called WARPEACE.TXT and
ANNAK.TXT, make sure that neither of these names
is ever used following the greater-than sign.
To check these correctly, you might do:
gutcheck warpeace.txt >war.lst
and
gutcheck annak.txt > annak.lst
separately. Then you can look at war.lst and annak.lst
to see the gutcheck reports.
* * * *
For existing 0.98 users upgrading to 0.99:
If you run on old 16-bit DOS or Windows 3.x, I'm afraid
you're out of luck. I'm not saying it _can't_ be compiled
to run on 16-bit, but the executable with the package is
for Win32 only. *nix users won't notice the change at all.
There are two new switches: -u and -d.
See above for full rundown.
Here's a list of the new errors:
Line 1456 - Carat character?
I^ve found a few.
Line 1821 - Forward slash?
Common error for italicized "I", or so /'ve found.
Line 2139 - Query missing paragraph break?
"Come here, son." "Do I _have_ to go, dad?"
Like that. False positives in some texts. Sorry 'bout that,
but these are often errors.
Line 2200 - Query had/bad error?
Clear enough. Doesn't catch as many as I'd like it to,
but rarely gives false alarms.
Line 2268 - Query punctuation after the?
Some words, like "the", very rarely have punctuation
following them. Others, like "Mrs", usually have a
period, but never a comma. Occasional false positives.
Line 2380 - Query possible scanno arid
It found one of your user-defined typos when you
used the -u switch.
Line 2511 - Capital "S"?
Surprisingly common specific case, like: Jane'S
Line 3469 - endquote missing punctuation?
OK. This one can really cause a lot of false positives
in some books, but it switches itself off if it finds
more than 20 in a text, unless you force it to list them
all with the -v switch.
"Hey, dad" Johnny said, "can we go now?"
is a common punctuation-missing error.
Line 4266 - Mismatched underscores?
Like mismatched anything else!