As is this script has limited purpose, and includes RTF markup not useful to anyone else other than as example.

But, what is useful is the date function that increments to the next ‘weekday’ day of week. Be aware that this script formats into Americanized short date (e.g., 12-31-69). Which may seem odd since I just wrote (and re-wrote a couple dozen times) a script to obliterate that format in file names. Except that I’m using this script to fill out the dates on an order form where the dates are expected to be the familiar, informal conversational format used in the US. Whereas for a file name in a computer, YYYY-MM-DD is a useful sorting method.

A lot of ‘others’ don’t take the time to realize that. I’ve run into vandalized man pages for date that attempt to make admonitions against the US informal conversational date format into some moral indictment. That format appears from the way one would speak the date in the US: December thirty-first, nineteen sixty-nine. The same as one would use more words to say: The thirty-first of December, nineteen sixty-nine. It’s not hard, Europeans. 😀

Also posting the plist used with launchd to run this script every morning.

#! /bin/sh
# http://strawhousepig.net/

# Used with launchd to run every morning user is logged in.
# Runs "at load" in case log in happens after the scheduled time (8:15).
# Purpose: $outfile is meant to be placed as a linked text object in InDesign document.
# But, InDesign (CS2) won't keep text style when updating the link
# unless you load it up with (double escaped) RTF markup. :|

# ProTip: Use '$todate' and '$nextdate' as placeholder text in your RTF file.
#         ie., format a sample RTF file then open, copy, & paste it as plain text here.

outfile=~/Documents/date-today.rtf

# Today's (formatted) date.
todate=$(date -j "+%m-%d-%y" | sed -E 's/0([0-9])/\1/g')

# Next weekday:
# Use today's day-of-week to count days until Monday if day-of-week is greater than 4 (Thursday). 
dofw=$(date +%w)
nextdate=$(date -j -v+$(( ( $dofw>4 )?8-$dofw:1))d "+%m-%d-%y" | sed -E 's/0([0-9])/\1/g')

rtf="{\\\rtf1\\\ansi\\\ansicpg1252\\\cocoartf1038\\\cocoasubrtf360
{\\\fonttbl\\\f0\\\fswiss\\\fcharset0 Helvetica;}
{\\\colortbl;\\\red255\\\green255\\\blue255;}
\\\margl1440\\\margr1440\\\vieww9000\\\viewh8400\\\viewkind0
\\\pard\\\tx720\\\tx1440\\\tx2160\\\tx2880\\\tx3600\\\tx4320\\\tx5040\\\tx5760\\\tx6480\\\tx7200\\\tx7920\\\tx8640\\\sb160\\\ql\\\qnatural\\\pardirnatural

\\\f0\\\fs32 \\\cf0 $todate\\\

$nextdate}"

printf "$rtf" > "$outfile"

Launchd properties file (reflects the above code being named “date-today-write-to-file.sh” and placed into the user’s ‘Documents’ folder) should be placed in the users Library folder, not the root Library (e.g., ~/Library/LaunchAgents/date-today.write-to-file.plist) and loaded with launchctl ~/Library/LaunchAgents/date-today.write-to-file.plist

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
  <dict>
    <key>Label</key>
    <string>date-today.write-to-file</string>
    <key>Program</key>
    <string>/Users/EXAMPLEUSER/Documents/date-today-write-to-file.sh</string>
    <key>RunAtLoad</key>
    <true/>
    <key>StartCalendarInterval</key>
    <dict>
      <key>Hour</key>
      <integer>08</integer>
      <key>Minute</key>
      <integer>15</integer>
    </dict>
  </dict>
</plist>

Annnnnd I’m done.

Did I say “done?” Now using a regex that will (should) only pick viable dates, although other strings can match. Notably a time string that meets the [01-12].[01-31].[00-99] format. Also added some error checking… or at least that’s the idea. Unsure if working at this time. I don’t believe it do. Only if `date` can’t reformat the string it’s fed will it be able to notice, but that is printed to stdout already. It’s the “Warning:” lines that are ambiguous.

Removed the “error checking” which didn’t work. I suggest making note of any errors reported to stdout while the script runs.

The main extended regex that matches the mm.dd.(yy)yy pattern is now a variable.

#!/bin/sh
# http://strawhousepig.net/

# The following `egrep` (or `grep -E`) should prove useful when checking the 
# storage file. If your text editor supports grep based 'find all' you 
# may be able to highlight the original text that is to be changed.
# egrep -nv '(1[0-2]|0?[1-9])[.-/,]([0-2]?[1-9]|[1-3][0-1])[.-/,](20)?[0-9]{2}(\.[[:alnum:]]+)?[[:space:]]/'

# Name of file in which to store 'mv' statements.
mvstore="mvstore.sh"

# The number of parentheses here affects the `sed` for $e. Currently "\4".
dateregex="(1[0-2]|0?[1-9])[.-/,]([0-2]?[1-9]|[1-3][0-1])[.-/,](20)?[0-9]{2}"

usage="
    This script looks recursively for the date pattern mm.dd.yy and similar
    in file and directory names beginning at 'pwd'. It then generates an 'mv'
    statement for each item with the pattern re-written by 'sed' & 'date'
    to yyyy-mm-dd.

    Running the '-e' option without first creating and checking the storage
    file is not recommended.

        -w    Write the 'mv' statements to the file '$mvstore'
                in the present working directory.
        -e    Evaluate the 'mv' statements as they are generated.
        -h    Displays this helpful text.
\n"

function mvdatef() {
    # Here, 'tail' reverses the order of 'find' after 'egrep' filters the result.
    # That way files in a directory are renamed before the directory is.
    # 'find' might be able to use the regex, but I couldn't work it out. YMMV.
    # This only looks for the pattern at the end of the line or just before an extension.
    # If something isn't working, the regex here is probably where it started.
    find "`pwd`" | egrep '[/[:space:]]'"$dateregex"'(\.[[:alnum:]]+)?$' | tail -r | while read a
    do

        # Escape certain characters so they don't wreck the 'mv' statement later.
        # double quote, single quote, parens, ampersand, dollar sign, and space.
        # Single quote, parens, and ampersand are escaped for the shell after breaking out of the 'sed' statement.
        b=$(echo "$a" | sed -E 's/(["'\'\(\)\&'[:space:]$])/\\\1/g')
        if [ "$b" == "" ]; then
            echo "Error: Could not escape $a" >> $mvstore
        fi
        # Suck out the last instance of our hated date pattern (00.00.00 or 00.00.0000).
        # Also replace dashes, slashes (not working :| ), and errant commas for dots
        # because we've come too far not to.
        c=$(echo "$a" | egrep -o "$dateregex" | tail -1 | sed -E 's/[-/,]/\./g')
        if [ "$c" != "" ]; then
            # 'date' will not accept a 2 OR 4 digit year.
            if [ $(echo $c | egrep -o "[0-9]{4}$") ]; then
                dform="%m.%d.%Y"; else
                dform="%m.%d.%y"
            fi
            d="$(date -j -f "$dform" "$c" "+%Y-%m-%d")"
            if [ "$d" == "" ]; then
                echo "Error: Could not format date from $c" >> $mvstore
            fi
            # This is the 'sed' that finds the date and replaces it with what we made just above.
            # It looks for the pattern at the end of the line (path) but includes the extension if there.
            # It is possible that a version number of some sort will also match.
            # Also possible to do away with the EOL in the regex and just go for the pattern.
            e="$(echo "$b" | sed -E 's/'"$dateregex"'(\.[[:alnum:]]+)?$/'$d'\4/')"
            if [ "$e" == "" ]; then
                echo "Error: Could not replace $c in $a" >> $mvstore
            fi
            # After all that dicking around, this is the mv statement.
            f="mv -- $b $e"

            if [ $1 ]; then
                eval $f
            else
                echo "$f" >> "$mvstore"
            fi
        else
            echo "Error: $c - Could not pull viable date from $b" >> $mvstore
        fi
    done
}

if [[ $1 = "-h" ]] ; then
    printf "$usage"
    exit
elif [[ $1 = "-e" ]]; then
    echo "Evaluating 'mv' statements as they are generated..."
    mvdatef -e
elif [[ $1 = "-w" ]]; then
    echo "Generating file `pwd`/$mvstore and writing 'mv' statements to it..."
    printf "#/bin/sh\n#  `date`\n" > "$mvstore"
    mvdatef
else
    printf "$usage"
    exit
fi

Original, less-good version: http://strawhousepig.net/shell-script-to-re-format-poor-date-format-in-filenames/

New, more-good version here: New script to re-write dates in file names

After testing and some using, this is going to change to just write the mv statements to a file and to accept an argument to run them instead. No logging since that will serve the same purpose. This post does not contain the final version. It’s here for the sake of history.

And boom. Or so I hope. This *should* handle dates as “01.02.03” and the equivalent “01.02.2003”.
To-do: Check for administrator privileges or just write the log to ~/. MacOS (nee OS X (nee Mac OS)) has a default user log directory, but GNU does not. I suppose we could check for “~/.log/” and mkdir if false…

Title explains it pretty well. Uses egrep to find date, sed to edit it, mv to rename, and read for user confirmation prompt. Which, btw, can’t be done if you are already doing a read without feeding the new read from somewhere other than stdin. Hence the < /dev/tty at that point in the script. So many strangers to thank for sharing their knowledge online. Thanks, strangers! That should do it.

Why this exists is I had a co-worker who added dates to tons of files in the format “01.02.03”, which resulted in files named “Example filename 01.02.03.xmpl” Same for phone numbers. Clearly this is wrong and something must be done about it (now that new ones won’t be cropping up since he.is.gone). You may be asking yourself how he got away with not getting an e-mailed file bounced back for multiple extensions? My educated guess is that he did and kept doing it the dumb way despite that.

Corrects the above format to “+%Y-%m-%d” (YYYY-MM-DD).

This could probably be done with fewer lines or overall be more betterer, but this is like my first whole shell script. Maybe my years of AppleScript are showing. *meh*
Continue reading →