diff --git a/experment/sweeps_to_note.md b/experment/sweeps_to_note.md
new file mode 100644
index 00000000..495d5e66
--- /dev/null
+++ b/experment/sweeps_to_note.md
@@ -0,0 +1 @@
+coveralls is to slow right now
diff --git a/src/dateformat.jl b/src/dateformat.jl
index fb09ebf3..13a2e9fa 100644
--- a/src/dateformat.jl
+++ b/src/dateformat.jl
@@ -3,8 +3,33 @@ Dates.default_format(::Type{NanoDate}) = ISONanoDateFormat
 
 NanoDate(nd::NanoDate, df::DateFormat) = format(nd, df)
 
-# returns the specifier part as a string
-Base.String(df::DateFormat) = string(df)[12:end-1]
+# get the specifier part of a DateFormat as a string
+# restricts the specifier to 9 subsecond digits
+function safestring(df::DateFormat)
+    str1 = string(df)[12:end-1]
+    str2 = ""
+    sepidx = findlast('.', str1)
+    isnothing(sepidx) && return str1
+    
+    if endswith(str1, "Z")
+        str1 = str1[1:end-1]
+        str2 = "Z"
+    elseif endswith(str1, "m") # "±hh:mm" or "±hhmm"
+        pmidx = findlast('h', str1)
+        if isnothing(pmidx) || pmidx == 1 || (str1[pmidx-1] != '+' || str1[pmidx-1] != '-')
+            return str1
+        else
+            pmidx -= 1
+            str1 = str1[1:pmidx-1]
+            str2 = str1[pmidx:end]
+        end
+    end
+    
+    sidxlast = findlast('s', str1)
+    sidxmax  = sepidx + 9
+    sidxlast = min(sidxlast, sidxmax)
+    str1[begin:sidxlast] * str2
+end
 
 omit(needle::Nothing, haystack::Nothing) = nothing
 omit(needle, haystack::Nothing) = nothing
@@ -51,7 +76,7 @@ function findeach(needle, haystack)
 end
 
 function nanodateformat(nd::NanoDate, df::DateFormat; subsecsep::Union{Char,AbstractString}='.')
-    dfstr = String(df)
+    dfstr = safestring(df)
     scount, sindices = findeach('s', dfstr)
     idxsubsecsep = subsecsep == "" ? nothing : findlast(subsecsep, dfstr)
     if !isnothing(idxsubsecsep)
diff --git a/src/strings.jl b/src/strings.jl
index b8e1c061..69794e19 100644
--- a/src/strings.jl
+++ b/src/strings.jl
@@ -82,7 +82,7 @@ charvec2string(chrs::AbstractVector{Char}) = foldl(*, chrs; init = "")
 function nanodate_format(nd, df)
     nooffset(df)
     indices = indexperiods(df)
-    dfstr = String(df)
+    dfstr = safestring(df)
     chrs = string2charvec(dfstr)
     indices = NamedTuple{(:yr, :mn, :dy, :hr, :mi, :sc, :ss)}(indices) # omit offset field
     syms = keys(indices)
@@ -142,7 +142,7 @@ end
     subsec_str = lpad(subsec_value, 9, '0')
     supersec = datetime - Millisecond(datetime)
     
-    dfstr = String(df)
+    dfstr = safestring(df)
     str = Dates.format(datetime, df)
     value(nd.nanosecs) == 0 && return str
 
@@ -195,7 +195,7 @@ function nanodate_format(nd, df, sep)
     str
 end
 
-nooffset(df::DateFormat) = nooffset(String(df))
+nooffset(df::DateFormat) = nooffset(safestring(df))
 function nooffset(str::AbstractString)
     if occursin('+', str)
         throw(ArgumentError("utc offsets are not supported in format(), use timestamp()"))
@@ -203,7 +203,7 @@ function nooffset(str::AbstractString)
 end
 
 
-separate_offset(df::DateFormat) = separate_offset(String(df))
+separate_offset(df::DateFormat) = separate_offset(safestring(df))
 
 function separate_offset(str::AbstractString)
     if isempty(str) || isdigit(str[end])
@@ -323,7 +323,7 @@ function getpart(r::UnitRange, str)
 end
 getpart(x::Nothing, str) = "0"
 
-indexperiods(df::DateFormat) = indexperiods(String(df))
+indexperiods(df::DateFormat) = indexperiods(safestring(df))
 
 function indexperiods(dfstr::String)
     str = strip(dfstr)